MLS-C01 AWS Certified Machine Learning - Specialty Questions and Answers

The best technique to improve the recall of the MLP for the target class of interest is to add class weights to the MLP’s loss function and then retrain. Class weights are a way of assigning different importance to each class in the dataset, such that the model will pay more attention to the classes with higher weights. This can help mitigate the class imbalance problem, where the model tends to favor the majority class and ignore the minority class. By increasing the weight of the target class of interest, the model will try to reduce the false negatives and increase the true positives, which will improve the recall metric. Adding class weights to the loss function is also a quick and easy solution, as it does not require gathering more data, changing the model architecture, or switching to a different algorithm.

AWS Machine Learning Specialty Exam Guide

AWS Machine Learning Training - Deep Learning with Amazon SageMaker

AWS Machine Learning Training - Class Imbalance and Weighted Loss Functions

The best visualization for this task is to create a bar plot, faceted by year, of average sales for each region and add a horizontal line in each facet to represent average sales. This way, the data scientist can easily compare the yearly average sales for each region with the overall average sales and see the trends over time. The bar plot also allows the data scientist to see the relative performance of each region within each year and across years. The other options are less effective because they either do not show the yearly trends, do not show the overall average sales, or do not group the data by region.

pandas.DataFrame.groupby — pandas 2.1.4 documentation

pandas.DataFrame.plot.bar — pandas 2.1.4 documentation

Matplotlib - Bar Plot - Online Tutorials Library

The solution that the agency should consider is to use a proxy server at each local office and for each camera, and stream the RTSP feed to a unique Amazon Kinesis Video Streams video stream. On each stream, use Amazon Rekognition Video and create a stream processor to detect faces from a collection of known employees, and alert when non-employees are detected.

This solution has the following advantages:

It can handle thousands of video cameras in real time, as Amazon Kinesis Video Streams can scale elastically to support any number of producers and consumers1.

It can leverage the Amazon Rekognition Video API, which is designed and optimized for video analysis, and can detect faces in challenging conditions such as low lighting, occlusions, and different poses2.

It can use a stream processor, which is a feature of Amazon Rekognition Video that allows you to create a persistent application that analyzes streaming video and stores the results in a Kinesis data stream3. The stream processor can compare the detected faces with a collection of known employees, which is a container for persisting faces that you want to search for in the input video stream4. The stream processor can also send notifications to Amazon Simple Notification Service (Amazon SNS) when non-employees are detected, which can trigger downstream actions such as sending alerts or storing the events in Amazon Elasticsearch Service (Amazon ES)3.

1: What Is Amazon Kinesis Video Streams? - Amazon Kinesis Video Streams

2: Detecting and Analyzing Faces - Amazon Rekognition

3: Using Amazon Rekognition Video Stream Processor - Amazon Rekognition

4: Working with Stored Faces - Amazon Rekognition

 The best solution to reduce the cost of the notebook instance and the data preprocessing job is to change the notebook instance type to a smaller general-purpose instance, stop the notebook when it is not in use, and run data preprocessing on an ml.r5 instance with the same memory size as the ml.m5.4xlarge instance by using Amazon SageMaker Processing. This solution will result in the most cost savings because:

Changing the notebook instance type to a smaller general-purpose instance will reduce the hourly cost of running the notebook, since the feature engineering development does not require high CPU and memory resources. For example, an ml.t3.medium instance costs $0.0464 per hour, while an ml.m5.4xlarge instance costs $0.888 per hour1.

Stopping the notebook when it is not in use will also reduce the cost, since the notebook will only incur charges when it is running. For example, if the notebook is used for 8 hours per day, 5 days per week, then stopping it when it is not in use will save about 76% of the monthly cost compared to leaving it running all the time2.

Running data preprocessing on an ml.r5 instance with the same memory size as the ml.m5.4xlarge instance by using Amazon SageMaker Processing will reduce the cost of the data preprocessing job, since the ml.r5 instance is optimized for memory-intensive workloads and has a lower cost per GB of memory than the ml.m5 instance. For example, an ml.r5.4xlarge instance has 128 GB of memory and costs $1.008 per hour, while an ml.m5.4xlarge instance has 64 GB of memory and costs $0.888 per hour1. Therefore, the ml.r5.4xlarge instance can process the same amount of data in half the time and at a lower cost than the ml.m5.4xlarge instance. Moreover, using Amazon SageMaker Processing will allow the data preprocessing job to run on a separate, fully managed infrastructure that can be scaled up or down as needed, without affecting the notebook instance.

The other options are not as effective as option C for the following reasons:

Option A is not optimal because changing the notebook instance type to a memory optimized instance with the same vCPU number as the ml.m5.4xlarge instance has will not reduce the cost of the notebook, since the memory optimized instances have a higher cost per vCPU than the general-purpose instances. For example, an ml.r5.4xlarge instance has 16 vCPUs and costs $1.008 per hour, while an ml.m5.4xlarge instance has 16 vCPUs and costs $0.888 per hour1. Moreover, running both data preprocessing and feature engineering development on the same instance will not take advantage of the scalability and flexibility of Amazon SageMaker Processing.

Option B is not suitable because running data preprocessing on a P3 instance type with the same memory as the ml.m5.4xlarge instance by using Amazon SageMaker Processing will not reduce the cost of the data preprocessing job, since the P3 instance type is optimized for GPU-based workloads and has a higher cost per GB of memory than the ml.m5 or ml.r5 instance types. For example, an ml.p3.2xlarge instance has 61 GB of memory and costs $3.06 per hour, while an ml.m5.4xlarge instance has 64 GB of memory and costs $0.888 per hour1. Moreover, the data preprocessing job does not require GPU, so using a P3 instance type will be wasteful and inefficient.

Option D is not feasible because running data preprocessing on an R5 instance with the same memory size as the ml.m5.4xlarge instance by using the Reserved Instance option will not reduce the cost of the data preprocessing job, since the Reserved Instance option requires a commitment to a consistent amount of usage for a period of 1 or 3 years3. However, the data preprocessing job only runs once a day on average and completes in only 2 hours, so it does not have a consistent or predictable usage pattern. Therefore, using the Reserved Instance option will not provide any cost savings and may incur additional charges for unused capacity.

Amazon SageMaker Pricing

Manage Notebook Instances - Amazon SageMaker

Amazon EC2 Pricing - Reserved Instances

 The TensorFlow estimator configuration that will train the model most cost-effectively is to turn on SageMaker Training Compiler by adding compiler_config=TrainingCompilerConfig() as a parameter, turn on managed spot training by setting the use_spot_instances parameter to True, and pass the script to the estimator in the call to the TensorFlow fit() method. This configuration will optimize the model for the target hardware platform, reduce the training cost by using Amazon EC2 Spot Instances, and use the custom Python script without any modification.

SageMaker Training Compiler is a feature of Amazon SageMaker that enables you to optimize your TensorFlow, PyTorch, and MXNet models for inference on a variety of target hardware platforms. SageMaker Training Compiler can improve the inference performance and reduce the inference cost of your models by applying various compilation techniques, such as operator fusion, quantization, pruning, and graph optimization. You can enable SageMaker Training Compiler by adding compiler_config=TrainingCompilerConfig() as a parameter to the TensorFlow estimator constructor1.

Managed spot training is another feature of Amazon SageMaker that enables you to use Amazon EC2 Spot Instances for training your machine learning models. Amazon EC2 Spot Instances let you take advantage of unused EC2 capacity in the AWS Cloud. Spot Instances are available at up to a 90% discount compared to On-Demand prices. You can use Spot Instances for various fault-tolerant and flexible applications. You can enable managed spot training by setting the use_spot_instances parameter to True and specifying the max_wait and max_run parameters in the TensorFlow estimator constructor2.

The TensorFlow estimator is a class in the SageMaker Python SDK that allows you to train and deploy TensorFlow models on SageMaker. You can use the TensorFlow estimator to run your own Python script on SageMaker, without any modification. You can pass the script to the estimator in the call to the TensorFlow fit() method, along with the location of your input data. The fit() method starts a SageMaker training job and runs your script as the entry point in the training containers3.

The other options are either less cost-effective or more complex to implement. Adjusting the training script to use distributed data parallelism would require modifying the script and specifying appropriate values for the distribution parameter, which could increase the development time and complexity. Setting the MaxWaitTimeInSeconds parameter to be equal to the MaxRuntimeInSeconds parameter would not reduce the cost, as it would only specify the maximum duration of the training job, regardless of the instance type.

1: Optimize TensorFlow, PyTorch, and MXNet models for deployment using Amazon SageMaker Training Compiler | AWS Machine Learning Blog

2: Managed Spot Training: Save Up to 90% On Your Amazon SageMaker Training Jobs | AWS Machine Learning Blog

3: sagemaker.tensorflow — sagemaker 2.66.0 documentation

 To use a custom algorithm in Amazon SageMaker, the Docker container image must have an executable file named train that acts as the ENTRYPOINT for the container. This file is responsible for running the training code and communicating with the Amazon SageMaker service. The train file must be in the PATH of the container and have execute permissions. The other options are not valid ways to package the Docker container for Amazon SageMaker. References:

Use Docker containers to build models - Amazon SageMaker

Create a container with your own algorithms and models - Amazon SageMaker

Feature selection is the process of reducing the number of input variables to those that are most relevant for predicting the target variable. One way to do this is to run a correlation check of all features against the target variable and remove features with low target variable correlation scores. This means that these features have little or no linear relationship with the target variable and are not useful for the prediction. This can reduce the model’s complexity and improve its performance. References:

Feature engineering - Machine Learning Lens

Feature Selection For Machine Learning in Python

 The solution that will meet the requirements with the least development time is to use Amazon Kinesis Data Firehose to stream the reservation data directly to Amazon S3, detect high-demand outliers by using Amazon QuickSight ML Insights, and visualize the data in QuickSight. This solution does not require any custom development or ML domain expertise, as it leverages the built-in features of QuickSight ML Insights to automatically run anomaly detection and generate insights on the streaming data. QuickSight ML Insights can also create a visualization dashboard that automatically refreshes with the most recent data, and allows the data scientist to explore the outliers and their key drivers. References:

1: Simplify and automate anomaly detection in streaming data with Amazon Lookout for Metrics | AWS Machine Learning Blog

2: Detecting outliers with ML-powered anomaly detection - Amazon QuickSight

3: Real-time Outlier Detection Over Streaming Data - IEEE Xplore

4: Towards a deep learning-based outlier detection … - Journal of Big Data

Amazon Kinesis Data Firehose is a service that can ingest streaming data and store it in various destinations, including Amazon S3, Amazon Redshift, Amazon Elasticsearch Service, and Splunk. Amazon Kinesis Data Firehose can also convert the incoming data to Apache Parquet or Apache ORC format before storing it in Amazon S3. This can reduce the storage cost and improve the performance of analytical queries on the data. Amazon Kinesis Data Firehose supports various data sources, such as Amazon Kinesis Data Streams, Amazon Managed Streaming for Apache Kafka, AWS IoT, and custom applications. Amazon Kinesis Data Firehose can also apply data transformation and compression using AWS Lambda functions.

AWSDMS is not a valid service name. AWS Database Migration Service (AWS DMS) is a service that can migrate data from various sources to various targets, but it does not support streaming data or Parquet format.

Amazon Kinesis Data Streams is a service that can ingest and process streaming data in real time, but it does not store the data in any destination. Amazon Kinesis Data Streams can be integrated with Amazon Kinesis Data Firehose to store the data in Parquet format.

Amazon Kinesis Data Analytics is a service that can analyze streaming data using SQL or Apache Flink, but it does not store the data in any destination. Amazon Kinesis Data Analytics can be integrated with Amazon Kinesis Data Firehose to store the data in Parquet format. References:

Amazon Kinesis Data Firehose - Amazon Web Services

What Is Amazon Kinesis Data Firehose? - Amazon Kinesis Data Firehose

Amazon Kinesis Data Firehose FAQs - Amazon Web Services

The solution that the data scientist should use to improve the performance of the model is to apply the Synthetic Minority Oversampling Technique (SMOTE) on the minority class in the training dataset, and retrain the model with the updated training data. This solution can address the problem of class imbalance in the dataset, which can affect the model’s ability to learn from the rare but important positive class (fraud).

Class imbalance is a common issue in machine learning, especially for classification tasks. It occurs when one class (usually the positive or target class) is significantly underrepresented in the dataset compared to the other class (usually the negative or non-target class). For example, in the credit card fraud detection problem, the positive class (fraud) is much less frequent than the negative class (fair transactions). This can cause the model to be biased towards the majority class, and fail to capture the characteristics and patterns of the minority class. As a result, the model may have a high overall accuracy, but a low recall or true positive rate for the minority class, which means it misses many fraudulent transactions.

SMOTE is a technique that can help mitigate the class imbalance problem by generating synthetic samples for the minority class. SMOTE works by finding the k-nearest neighbors of each minority class instance, and randomly creating new instances along the line segments connecting them. This way, SMOTE can increase the number and diversity of the minority class instances, without duplicating or losing any information. By applying SMOTE on the minority class in the training dataset, the data scientist can balance the classes and improve the model’s performance on the positive class1.

The other options are either ineffective or counterproductive. Applying SMOTE on the majority class would not balance the classes, but increase the imbalance and the size of the dataset. Undersampling the minority class would reduce the number of instances available for the model to learn from, and potentially lose some important information. Oversampling the majority class would also increase the imbalance and the size of the dataset, and introduce redundancy and overfitting.

1: SMOTE for Imbalanced Classification with Python - Machine Learning Mastery

The best option is to use Amazon SageMaker Studio to rebuild the model and deploy it at an endpoint. Then, use Amazon SageMaker Model Monitor to store inferences and use the inferences to create Shapley values that help explain model behavior. Shapley values are a way of attributing the contribution of each feature to the model output. They can help the credit team understand why the model makes certain decisions and how the features affect the model outcomes. A chart that shows features and SHapley Additive exPlanations (SHAP) values can be created using the SHAP library in Python. This option is the most operationally efficient because it leverages the existing XGBoost training container and the built-in capabilities of Amazon SageMaker Model Monitor and SHAP library. References:

Amazon SageMaker Studio

Amazon SageMaker Model Monitor

SHAP library

The data scientist should use Amazon SageMaker Ground Truth to sort the data into two groups named “enrolled” or “not enrolled.” This will create a labeled dataset that can be used for supervised learning. The data scientist should then use a classification algorithm to run predictions on the test data. A classification algorithm is a suitable choice for predicting a binary outcome, such as enrollment status, based on the input features, such as academic performance. A classification algorithm will output a probability for each class label and assign the most likely label to each observation.

Use Amazon SageMaker Ground Truth to Label Data

Classification Algorithm in Machine Learning

Stratified sampling is a technique that preserves the class distribution of the original dataset when creating a smaller or split dataset. This means that the proportion of examples from each class in the original dataset is maintained in the smaller or split dataset. Stratified sampling can help improve the validation accuracy of the model by ensuring that the validation dataset is representative of the original dataset and not biased towards any class. This can reduce the variance and overfitting of the model and increase its generalization ability. Stratified sampling can be applied to both oversampling and undersampling methods, depending on whether the goal is to increase or decrease the size of the dataset.

The other options are not effective ways to improve the validation accuracy of the model. Acquiring additional data about the majority classes in the original dataset will only increase the imbalance and make the model more biased towards the majority classes. Using a smaller, randomly sampled version of the training dataset will not guarantee that the class distribution is preserved and may result in losing important information from the minority classes. Performing systematic sampling on the original dataset will also not ensure that the class distribution is preserved and may introduce sampling bias if the original dataset is ordered or grouped by class.

Amazon SageMaker script mode allows users to bring custom training scripts (such as train.py) without needing extensive modifications for specific data formats like TFRecord. By storing the TFRecord data in an Amazon S3 bucket and pointing the SageMaker training job to this bucket, the model can directly access the data, allowing the ML specialist to train the model without additional reformatting or data processing steps.

This approach minimizes development overhead and leverages SageMaker's built-in support for custom training scripts and S3 integration, making it the most efficient choice.

 Amazon SageMaker is a service that allows users to build, train, and deploy ML models on AWS. Amazon SageMaker endpoints are scalable and secure web services that can be used to perform real-time inference on ML models. An endpoint configuration defines the models that are deployed and the resources that are used by the endpoint. An endpoint configuration can have multiple production variants, each representing a different version or variant of a model. Users can specify the portion of the inferences served by each production variant using the initialVariantWeight parameter. Users can also programmatically update the endpoint configuration to change the portion of the inferences served by each production variant using the UpdateEndpointWeightsAndCapacities API. Therefore, option B is the best solution to satisfy the requirements with minimal effort.

Option A is incorrect because creating multiple endpoints for each model would incur more cost and complexity than using a single endpoint with multiple production variants. Moreover, controlling the invocation of different models at the application layer would require more custom logic and coordination than using the UpdateEndpointWeightsAndCapacities API. Option C is incorrect because Amazon SageMaker Neo is a service that allows users to optimize ML models for different hardware platforms, such as edge devices. It is not relevant to the problem of running multiple versions of a model in parallel for long periods of time. Option D is incorrect because Amazon SageMaker batch transform is a service that allows users to perform asynchronous inference on large datasets. It is not suitable for the problem of performing real-time inference on streaming data from device users.

Deploying models to Amazon SageMaker hosting services - Amazon SageMaker

Update an Amazon SageMaker endpoint to accommodate new models - Amazon SageMaker

UpdateEndpointWeightsAndCapacities - Amazon SageMaker

The services that are integrated with Amazon SageMaker to track the information that the Machine Learning Specialist needs are AWS CloudTrail and Amazon CloudWatch. AWS CloudTrail is a service that records the API calls and events for AWS services, including Amazon SageMaker. AWS CloudTrail can track the actions performed by the Data Scientists, such as creating notebooks, training models, and deploying endpoints. AWS CloudTrail can also provide information such as the identity of the user, the time of the action, the parameters used, and the response elements returned. AWS CloudTrail can help the Machine Learning Specialist to monitor the usage and activity of Amazon SageMaker, as well as to audit and troubleshoot any issues1 Amazon CloudWatch is a service that collects and analyzes the metrics and logs for AWS services, including Amazon SageMaker. Amazon CloudWatch can track the performance and utilization of the Amazon SageMaker endpoints, such as the CPU and GPU utilization, the inference latency, the number of invocations, etc. Amazon CloudWatch can also track the errors and alarms that are generated when an endpoint is invoked, such as the model errors, the throttling errors, the HTTP errors, etc. Amazon CloudWatch can help the Machine Learning Specialist to optimize the operational performance and reliability of Amazon SageMaker, as well as to set up notifications and actions based on the metrics and logs

To log Amazon SageMaker API calls, the team can use AWS CloudTrail, which is a service that provides a record of actions taken by a user, role, or an AWS service in SageMaker1. CloudTrail captures all API calls for SageMaker, with the exception of InvokeEndpoint and InvokeEndpointAsync, as events1. The calls captured include calls from the SageMaker console and code calls to the SageMaker API operations1. The team can create a trail to enable continuous delivery of CloudTrail events to an Amazon S3 bucket, and configure other AWS services to further analyze and act upon the event data collected in CloudTrail logs1. The auditors can view the CloudTrail log activity report in the CloudTrail console or download the log files from the S3 bucket1.

To receive a notification when the model is overfitting, the team can add code to push a custom metric to Amazon CloudWatch, which is a service that provides monitoring and observability for AWS resources and applications2. The team can use the MXNet metric API to define and compute the custom metric, such as the validation accuracy or the validation loss, and use the boto3 CloudWatch client to put the metric data to CloudWatch3 . The team can then create an alarm in CloudWatch with Amazon SNS to receive a notification when the custom metric crosses a threshold that indicates overfitting . For example, the team can set the alarm to trigger when the validation loss increases for a certain number of consecutive periods, which means the model is learning the noise in the training data and not generalizing well to the validation data.

1: Log Amazon SageMaker API Calls with AWS CloudTrail - Amazon SageMaker

2: What Is Amazon CloudWatch? - Amazon CloudWatch

3: Metric API — Apache MXNet documentation

CloudWatch — Boto 3 Docs 1.20.21 documentation

Creating Amazon CloudWatch Alarms - Amazon CloudWatch

What is Amazon Simple Notification Service? - Amazon Simple Notification Service

Overfitting and Underfitting - Machine Learning Crash Course

The solution A will create the required transformed records with the least operational overhead because it uses AWS Lambda and Amazon Kinesis Data Firehose, which are fully managed services that can provide the desired functionality. The solution A involves the following steps:

Create an AWS Lambda function that can transform the incoming records. AWS Lambda is a service that can run code without provisioning or managing servers. AWS Lambda can execute the transformation logic on the purchasing records and add the new attributes to the records1.

Enable data transformation on the ingestion Kinesis Data Firehose delivery stream. Use the Lambda function as the invocation target. Amazon Kinesis Data Firehose is a service that can capture, transform, and load streaming data into AWS data stores. Amazon Kinesis Data Firehose can enable data transformation and invoke the Lambda function to process the incoming records before delivering them to Amazon S3. This can reduce the operational overhead of managing the transformation process and the data storage2.

The other options are not suitable because:

Option B: Deploying an Amazon EMR cluster that runs Apache Spark and includes the transformation logic, using Amazon EventBridge (Amazon CloudWatch Events) to schedule an AWS Lambda function to launch the cluster each day and transform the records that accumulate in Amazon S3, and delivering the transformed records to Amazon S3 will incur more operational overhead than using AWS Lambda and Amazon Kinesis Data Firehose. The company will have to manage the Amazon EMR cluster, the Apache Spark application, the AWS Lambda function, and the Amazon EventBridge rule. Moreover, this solution will introduce a delay in the transformation process, as it will run only once a day3.

Option C: Deploying an Amazon S3 File Gateway in the stores, updating the in-store software to deliver data to the S3 File Gateway, and using a scheduled daily AWS Glue job to transform the data that the S3 File Gateway delivers to Amazon S3 will incur more operational overhead than using AWS Lambda and Amazon Kinesis Data Firehose. The company will have to manage the S3 File Gateway, the in-store software, and the AWS Glue job. Moreover, this solution will introduce a delay in the transformation process, as it will run only once a day4.

Option D: Launching a fleet of Amazon EC2 instances that include the transformation logic, configuring the EC2 instances with a daily cron job to transform the records that accumulate in Amazon S3, and delivering the transformed records to Amazon S3 will incur more operational overhead than using AWS Lambda and Amazon Kinesis Data Firehose. The company will have to manage the EC2 instances, the transformation code, and the cron job. Moreover, this solution will introduce a delay in the transformation process, as it will run only once a day5.

1: AWS Lambda

2: Amazon Kinesis Data Firehose

3: Amazon EMR

4: Amazon S3 File Gateway

5: Amazon EC2

To compare a custom scikit-learn model with a benchmark model in a managed environment, the most effective and maintainable solution is to extend an existing SageMaker scikit-learn container.

“If you are using a framework like scikit-learn and need custom logic, you can extend the prebuilt SageMaker containers to include your own inference or training scripts. AWS manages the container base and you only maintain your code.”

This approach allows the data scientist to maintain control over model logic while letting SageMaker handle the container lifecycle, scaling, and infrastructure management—which aligns with the requirement for AWS to handle operational maintenance.

Multiple imputation is a technique that uses machine learning to generate multiple plausible values for each missing value in a dataset, based on the observed data and the relationships among the variables. Multiple imputation preserves the integrity of the dataset by accounting for the uncertainty and variability of the missing data, and avoids the bias and loss of information that may result from other methods, such as listwise deletion, last observation carried forward, or mean substitution. Multiple imputation can improve the accuracy and validity of statistical analysis and machine learning models that use the imputed dataset. References:

Managing missing values in your target and related datasets with automated imputation support in Amazon Forecast

Imputation by feature importance (IBFI): A methodology to impute missing data in large datasets

Multiple Imputation by Chained Equations (MICE) Explained

The data scientist should use K-means and principal component analysis (PCA) to meet this requirement. K-means is a clustering algorithm that can group customers based on their similarity in the feature space. PCA is a dimensionality reduction technique that can transform the original 980 variables into a smaller set of uncorrelated variables that capture most of the variance in the data. This can help reduce the computational cost and noise in the data, and improve the performance of the clustering algorithm.

Clustering - Amazon SageMaker

Dimensionality Reduction - Amazon SageMaker

In predictive maintenance, when a dataset is imbalanced (with far fewer failure cases than non-failure cases), oversampling the minority class helps the model learn from the minority class effectively. The Synthetic Minority Oversampling Technique (SMOTE) generates synthetic samples for the minority class by creating data points between existing minority class instances. This can enhance the model’s ability to recognize failure patterns, particularly in imbalanced datasets.

SMOTE increases the effective presence of failure cases in the dataset, providing a balanced learning environment for the model. This is more effective than undersampling, which would risk losing important non-failure data.

The solution that will meet the requirements with the least amount of customization to transform and store the ingested data is to use Amazon Kinesis Data Analytics to read and aggregate the data hourly, transform the data and store it in Amazon S3 by using Amazon Kinesis Data Firehose. This solution leverages the built-in features of Kinesis Data Analytics to perform SQL queries on streaming data and generate hourly summaries. Kinesis Data Analytics can also output the transformed data to Kinesis Data Firehose, which can then deliver the data to S3 in a specified format and partitioning scheme. This solution does not require any custom code or additional infrastructure to process the data. The other solutions either require more customization (such as using Lambda or EMR) or do not meet the requirement of aggregating the data hourly (such as using Lambda to read the data from Kinesis Data Streams). References:

1: Boosting Resiliency with an ML-based Telemetry Analytics Architecture | AWS Architecture Blog

2: AWS Cloud Data Ingestion Patterns and Practices

3: IoT ingestion and Machine Learning analytics pipeline with AWS IoT …

4: AWS IoT Data Ingestion Simplified 101: The Complete Guide - Hevo Data

 The best way to securely access data stored in a specific Amazon S3 bucket from an Amazon SageMaker notebook instance is to attach a policy to the IAM role associated with the notebook that allows GetObject, PutObject, and ListBucket operations to the specific S3 bucket. This way, the notebook can use the AWS SDK or CLI to access the S3 bucket without exposing any credentials or requiring any additional configuration. This is also the recommended approach by AWS for granting access to S3 from SageMaker. References:

Amazon SageMaker Roles

Accessing Amazon S3 from a SageMaker Notebook Instance

The problem of poor performance on the test data is a sign of overfitting, which means the model has learned the training data too well and failed to generalize to new and unseen data. To correct this, the Machine Learning Specialist should consider using methods that reduce the complexity of the model and increase its ability to generalize. Some of these methods are:

Increase regularization: Regularization is a technique that adds a penalty term to the loss function of the model, which reduces the magnitude of the model weights and prevents overfitting. There are different types of regularization, such as L1, L2, and elastic net, that apply different penalties to the weights1.

Increase dropout: Dropout is a technique that randomly drops out some units or connections in the neural network during training, which reduces the co-dependency of the units and prevents overfitting. Dropout can be applied to different layers of the network, and the dropout rate can be tuned to control the amount of dropout2.

Decrease feature combinations: Feature combinations are the interactions between different input features that can be used to create new features for the model. However, too many feature combinations can increase the complexity of the model and cause overfitting. Therefore, the Specialist should decrease the number of feature combinations and select only the most relevant and informative ones for the model3.

1: Regularization for Deep Learning - Amazon SageMaker

2: Dropout - Amazon SageMaker

3: Feature Engineering - Amazon SageMaker

To retrieve only the latest version of a record in real-time for a single entity (like a customer), the correct method is to use the GetRecord API provided by SageMaker Feature Store.

“Use the GetRecord API to retrieve the latest feature values for a specific record identifier from the online store. This operation returns the latest version of the record only.”

This is optimal for low-latency, single-record lookups, such as real-time inference scenarios. BatchGetRecord is used for multiple records, and Athena queries are for offline/batch analysis and not suited for real-time access.

The best way to maintain the privacy and integrity of the data stored in Amazon S3 is to use a combination of VPC endpoints and S3 bucket policies. A VPC endpoint allows the SageMaker notebook instance to access the S3 bucket without going through the public internet. A bucket policy allows the S3 bucket owner to specify which VPCs or VPC endpoints can access the bucket. This way, the data is protected from unauthorized access and tampering. The other options are either insecure (A and D) or inefficient (B). References: Using Amazon S3 VPC Endpoints, Using Bucket Policies and User Policies

The best way to fix the problem without changing the Lambda code or data in DynamoDB is to add the unrecognized words as synonyms in the custom slot type. This way, Amazon Lex can resolve the synonyms to the corresponding slot values and pass them to the Lambda function. For example, if the slot type has a value “comedy” with synonyms “funny”, “fun”, and “humor”, then any of these words entered by the user will be resolved to “comedy” and the Lambda function can query the DynamoDB table for the book titles in that category. Adding synonyms to the custom slot type can be done easily using the Amazon Lex console or API, and does not require any code changes.

The other options are not correct because:

Option A: Adding the unrecognized words in the enumeration values list as new values in the slot type would not fix the problem, because the Lambda function and the DynamoDB table are not aware of these new values. The Lambda function would not be able to query the DynamoDB table for the book titles in the new categories, and the request would still fail. Moreover, adding new values to the slot type would increase the complexity and maintenance of the chatbot, as the Lambda function and the DynamoDB table would have to be updated accordingly.

Option B: Creating a new custom slot type, adding the unrecognized words to this slot type as enumeration values, and using this slot type for the slot would also not fix the problem, for the same reasons as option A. The Lambda function and the DynamoDB table would not be able to handle the new slot type and its values, and the request would still fail. Furthermore, creating a new slot type would require more effort and time than adding synonyms to the existing slot type.

Option C: Using the AMAZON.SearchQuery built-in slot types for custom searches in the database is not a suitable approach for this use case. The AMAZON.SearchQuery slot type is used to capture free-form user input that corresponds to a search query. However, this slot type does not perform any validation or resolution of the user input, and passes the raw input to the Lambda function. This means that the Lambda function would have to handle the logic of parsing and matching the user input to the DynamoDB table, which would require changing the Lambda code and adding more complexity to the solution.

Custom slot type - Amazon Lex

Using Synonyms - Amazon Lex

Built-in Slot Types - Amazon Lex

 Amazon Forecast requires the input data to be in a specific format. The data scientist should use ETL jobs in AWS Glue to separate the dataset into a target time series dataset and an item metadata dataset. The target time series dataset should contain the timestamp, item_id, and demand columns, while the item metadata dataset should contain the item_id, category, and lead_time columns. Both datasets should be uploaded as .csv files to Amazon S3 . References:

How Amazon Forecast Works - Amazon Forecast

Choosing Datasets - Amazon Forecast

 The issue described in the question is a sign of overfitting, which is a common problem in machine learning when the model learns the noise and details of the training data too well and fails to generalize to new and unseen data. Overfitting can result in a low training error rate but a high test error rate, which indicates poor performance and validity of the model. There are several techniques that can be used to prevent or reduce overfitting, such as data augmentation and regularization.

Data augmentation is a technique that applies various transformations to the original training data, such as rotation, scaling, cropping, flipping, adding noise, changing brightness, etc., to create new and diverse data samples. Data augmentation can increase the size and diversity of the training data, which can help the model learn more features and patterns and reduce the variance of the model. Data augmentation is especially useful for image data, as it can simulate different scenarios and perspectives that the model may encounter in real life. For example, in the question, the device uses a camera to observe drivers’ behavior, so data augmentation can help the model deal with different lighting conditions, angles, distances, etc. Data augmentation can be done using various libraries and frameworks, such as TensorFlow, PyTorch, Keras, OpenCV, etc12

Regularization is a technique that adds a penalty term to the model’s objective function, which is typically based on the model’s parameters. Regularization can reduce the complexity and flexibility of the model, which can prevent overfitting by avoiding learning the noise and details of the training data. Regularization can also improve the stability and robustness of the model, as it can reduce the sensitivity of the model to small fluctuations in the data. There are different types of regularization, such as L1, L2, dropout, etc., but they all have the same goal of reducing overfitting. L2 regularization, also known as weight decay or ridge regression, is one of the most common and effective regularization techniques. L2 regularization adds the squared norm of the model’s parameters multiplied by a regularization parameter (lambda) to the model’s objective function. L2 regularization can shrink the model’s parameters towards zero, which can reduce the variance of the model and improve the generalization ability of the model. L2 regularization can be implemented using various libraries and frameworks, such as TensorFlow, PyTorch, Keras, Scikit-learn, etc34

The other options are not valid or relevant for resolving the issue of overfitting. Adding vanishing gradient to the model is not a technique, but a problem that occurs when the gradient of the model’s objective function becomes very small and the model stops learning. Making the neural network architecture complex is not a solution, but a possible cause of overfitting, as a complex model can have more parameters and more flexibility to fit the training data too well. Using gradient checking in the model is not a technique, but a debugging method that verifies the correctness of the gradient computation in the model. Gradient checking is not related to overfitting, but to the implementation of the model.

Amazon SageMaker supports encryption at rest for the ML storage volumes, the model artifacts, and the data in Amazon S3 using AWS Key Management Service (AWS KMS). AWS KMS is a service that allows customers to create and manage encryption keys that can be used to encrypt data. AWS KMS also provides an audit trail of key usage by logging key events to AWS CloudTrail. Customers can use either AWS managed keys or customer managed keys to encrypt their data. AWS managed keys are created and managed by AWS on behalf of the customer, while customer managed keys are created and managed by the customer. Customer managed keys offer more control and flexibility over the key policies, permissions, and rotation. Therefore, to meet the requirements of the company, the best option is to use customer managed keys in AWS KMS to encrypt the ML data volumes, and to encrypt the model artifacts and data in Amazon S3.

The other options are not correct because:

Option A: AWS Cloud HSM is a service that provides hardware security modules (HSMs) to store and use encryption keys. AWS Cloud HSM is not integrated with Amazon SageMaker, and cannot be used to encrypt the ML data volumes, the model artifacts, or the data in Amazon S3. AWS Cloud HSM is more suitable for customers who need to meet strict compliance requirements or who need direct control over the HSMs.

Option B: SageMaker built-in transient keys are temporary keys that are used to encrypt the ML data volumes and are discarded immediately after encryption. These keys do not provide persistent encryption or logging of key usage. Enabling default encryption for new Amazon Elastic Block Store (Amazon EBS) volumes does not affect the ML data volumes, which are encrypted separately by SageMaker. Moreover, this option does not address the encryption of the model artifacts and data in Amazon S3.

Option D: AWS Security Token Service (AWS STS) is a service that provides temporary credentials to access AWS resources. AWS STS does not provide encryption keys or encryption services. AWS STS cannot be used to encrypt the ML storage volumes, the model artifacts, or the data in Amazon S3.

Protect Data at Rest Using Encryption - Amazon SageMaker

What is AWS Key Management Service? - AWS Key Management Service

What is AWS CloudHSM? - AWS CloudHSM

What is AWS Security Token Service? - AWS Security Token Service

Amazon SageMaker’s Neural Topic Model (NTM) is designed to uncover underlying topics within text data by clustering documents based on topic similarity. For document categorization, NTM can identify product categories by analyzing and grouping the documents, making it an efficient choice for unsupervised learning where predefined categories do not exist.

This model minimizes operational overhead by allowing automated topic categorization directly within SageMaker, which is more straightforward than custom Docker image setups or using k-means clustering for tokenized data.

The data scientist should obtain a correlated equilibrium policy by formulating this problem as a multi-agent reinforcement learning problem. This is because:

Multi-agent reinforcement learning (MARL) is a subfield of reinforcement learning that deals with learning and coordination of multiple agents that interact with each other and the environment 1. MARL can be applied to problems that involve distributed decision making, such as traffic signal control, where each traffic light can be modeled as an agent that observes the traffic state and chooses an action (e.g., changing the signal phase) to optimize a reward function (e.g., minimizing the delay or congestion) 2.

A correlated equilibrium is a solution concept in game theory that generalizes the notion of Nash equilibrium. It is a probability distribution over the joint actions of the agents that satisfies the following condition: no agent can improve its expected payoff by deviating from the distribution, given that it knows the distribution and the actions of the other agents 3. A correlated equilibrium can capture the correlation among the agents’ actions, which is useful for modeling the traffic behavior at each light that is subject to a small stochastic error term.

A correlated equilibrium policy is a policy that induces a correlated equilibrium in a MARL setting. It can be obtained by using various methods, such as policy gradient, actor-critic, or Q-learning algorithms, that can learn from the feedback of the environment and the communication among the agents 4. A correlated equilibrium policy can achieve a better performance than a Nash equilibrium policy, which assumes that the agents act independently and ignore the correlation among their actions 5.

Therefore, by obtaining a correlated equilibrium policy by formulating this problem as a MARL problem, the data scientist can most effectively model the traffic behavior and reduce congestion.

Multi-Agent Reinforcement Learning

Multi-Agent Reinforcement Learning for Traffic Signal Control: A Survey

Correlated Equilibrium

Multi-Agent Actor-Critic for Mixed Cooperative-Competitive Environments

Correlated Q-Learning

Based on the figure provided, the data is not linearly separable. Therefore, a non-linear model such as SVM with a non-linear kernel would be the best choice. SVMs are particularly effective in high-dimensional spaces and are versatile in that they can be used for both linear and non-linear data. Additionally, SVMs have a high level of accuracy and are less prone to overfitting1

 1: https://docs.aws.amazon.com/sagemaker/latest/dg/svm.html

The best way to retrain the model to meet the requirements is to set the target_recall hyperparameter to 90% and set the binary_classifier_model_selection_criteria hyperparameter to recall_at_target_precision. This will instruct the linear learner algorithm to optimize the model for a high recall score, while maintaining a reasonable precision score. Recall is the proportion of actual positives that were identified correctly, which is important for the company’s goal of reaching at least 90% of the customers who are likely to buy the new product1. Precision is the proportion of positive identifications that were actually correct, which is also relevant for the company’s budget and efficiency2. By setting the target_recall to 90%, the algorithm will try to achieve a recall score of at least 90%, and by setting the binary_classifier_model_selection_criteria to recall_at_target_precision, the algorithm will select the model that has the highest recall score among those that have a precision score equal to or higher than the target precision3. The target precision is automatically set to the median of the precision scores of all the models trained in parallel4.

The other options are not correct or optimal, because they have the following drawbacks:

B: Setting the target_precision hyperparameter to 90% and setting the binary_classifier_model_selection_criteria hyperparameter to precision_at_target_recall will optimize the model for a high precision score, while maintaining a reasonable recall score. However, this is not aligned with the company’s goal of reaching at least 90% of the customers who are likely to buy the new product, as precision does not reflect how well the model identifies the actual positives1. Moreover, setting the target_precision to 90% might be too high and unrealistic for the dataset, as the current precision score is only 75%4.

C: Using 90% of the historical data for training and setting the number of epochs to 20 will not necessarily improve the recall score of the model, as it does not change the optimization objective or the model selection criteria. Moreover, using more data for training might reduce the amount of data available for validation, which is needed for selecting the best model among the ones trained in parallel3. The number of epochs is also not a decisive factor for the recall score, as it depends on the learning rate, the optimizer, and the convergence of the algorithm5.

D: Setting the normalize_label hyperparameter to true and setting the number of classes to 2 will not affect the recall score of the model, as these are irrelevant hyperparameters for binary classification problems. The normalize_label hyperparameter is only applicable for regression problems, as it controls whether the label is normalized to have zero mean and unit variance3. The number of classes hyperparameter is only applicable for multiclass classification problems, as it specifies the number of output classes3.

1: Classification: Precision and Recall | Machine Learning | Google for Developers

2: Precision and recall - Wikipedia

3: Linear Learner Algorithm - Amazon SageMaker

4: How linear learner works - Amazon SageMaker

5: Getting hands-on with Amazon SageMaker Linear Learner - Pluralsight

Recall and precision are two metrics that can be used to evaluate the performance of a classification model. Recall is the ratio of true positives to the total number of actual positives, which measures how well the model can identify all the relevant cases. Precision is the ratio of true positives to the total number of predicted positives, which measures how accurate the model is when it makes a positive prediction. Based on the confusion matrix in the image, we can calculate the recall and precision as follows:

Recall = TP / (TP + FN) = 12 / (12 + 1) = 0.92

Precision = TP / (TP + FP) = 12 / (12 + 3) = 0.8

Where TP is the number of true positives, FN is the number of false negatives, and FP is the number of false positives. Therefore, the recall and precision of the model are 0.92 and 0.8, respectively.

Forecasting is a type of machine learning model that predicts future values of a target variable based on historical data and other features. Forecasting is suitable for problems that involve time-series data, such as the number of claims in each category from month to month. Forecasting can handle multiple categories of the target variable, as well as missing or partial information on some features. Therefore, option C is the best choice for the given problem.

Option A is incorrect because classification is a type of machine learning model that assigns a label to an input based on predefined categories. Classification is not suitable for predicting continuous or numerical values, such as the number of claims in each category from month to month. Moreover, classification requires sufficient and complete information on the features that are relevant to the target variable, which is not the case for the given problem. Option B is incorrect because reinforcement learning is a type of machine learning model that learns from its own actions and rewards in an interactive environment. Reinforcement learning is not suitable for problems that involve historical data and do not require an agent to take actions. Option D is incorrect because it combines two different types of machine learning models, which is unnecessary and inefficient. Moreover, classification is not suitable for predicting the number of claims in some categories, as explained in option A.

Forecasting | AWS Solutions for Machine Learning (AI/ML) | AWS Solutions Library

Time Series Forecasting Service – Amazon Forecast – Amazon Web Services

Amazon Forecast: Guide to Predicting Future Outcomes - Onica

Amazon Launches What-If Analyses for Machine Learning Forecasting …

A private subnet is a subnet that does not have a route to the internet gateway, which means that the resources in the private subnet cannot access the internet or be accessed from the internet. An S3 VPC endpoint is a gateway endpoint that allows the resources in the VPC to access the S3 service without going through the internet. By launching the notebook instances in a private subnet and accessing the data through an S3 VPC endpoint, the company can set up the job in a secure and compliant way, as the data never leaves the AWS network and is not exposed to the internet. This can also improve the performance and reliability of the data transfer, as the traffic does not depend on the internet bandwidth or availability.

Amazon VPC Endpoints - Amazon Virtual Private Cloud

Endpoints for Amazon S3 - Amazon Virtual Private Cloud

Connect to SageMaker Within your VPC - Amazon SageMaker

Working with VPCs and Subnets - Amazon Virtual Private Cloud

A hyperparameter tuning job is a feature of Amazon SageMaker that allows automatically finding the best combination of hyperparameters for a machine learning model. Hyperparameters are high-level parameters that influence the learning process and the performance of the model, such as the learning rate, the number of layers, the regularization factor, etc. A hyperparameter tuning job works by launching multiple training jobs with different hyperparameters, evaluating the results using an objective metric, and choosing the next set of hyperparameters to try based on a search strategy. The objective metric is a measure of the quality of the model, such as accuracy, precision, recall, etc. The search strategy is a method of exploring the hyperparameter space, such as random search, grid search, or Bayesian optimization.

Among the four options, option C is the most repeatable and requires the least amount of effort to use hyperparameter optimization to increase the model’s accuracy. This option involves the following steps:

Create a hyperparameter tuning job: Amazon SageMaker provides an easy-to-use interface for creating a hyperparameter tuning job, either through the AWS Management Console, the AWS CLI, or the AWS SDKs. To create a hyperparameter tuning job, the Machine Learning Specialist needs to specify the following information:

The name and type of the algorithm to use, either a built-in algorithm or a custom algorithm.

The ranges and types of the hyperparameters to tune, such as categorical, continuous, or integer.

The name and type of the objective metric to optimize, such as accuracy, and whether to maximize or minimize it.

The resource limits for the tuning job, such as the maximum number of training jobs and the maximum parallel training jobs.

The input data channels and the output data location for the training jobs.

The configuration of the training instances, such as the instance type, the instance count, the volume size, etc.

Set the accuracy as an objective metric: To use accuracy as an objective metric, the Machine Learning Specialist needs to ensure that the training algorithm writes the accuracy value to a file called metric_definitions in JSON format and prints it to stdout or stderr. For example, the file can contain the following content:

This means that the training algorithm prints a line like this:

Amazon SageMaker reads the accuracy value from the line and uses it to evaluate and compare the training jobs.

The other options are not as repeatable and require more effort than option C for the following reasons:

Option A: This option requires manually launching multiple training jobs in parallel with different hyperparameters, which can be tedious and error-prone. It also requires manually monitoring and comparing the results of the training jobs, which can be time-consuming and subjective.

Option B: This option requires writing code to create an AWS Step Functions workflow that monitors the accuracy in Amazon CloudWatch Logs and relaunches the training job with a defined list of hyperparameters, which can be complex and challenging. It also requires maintaining and updating the list of hyperparameters, which can be inefficient and suboptimal.

Option D: This option requires writing code to create a random walk in the parameter space to iterate through a range of values that should be used for each individual hyperparameter, which can be unreliable and unpredictable. It also requires defining and implementing a stopping criterion, which can be arbitrary and inconsistent.

Automatic Model Tuning - Amazon SageMaker

Define Metrics to Monitor Model Performance

The prior probability distribution for the discrete random variable that represents the number of minutes New Yorkers wait for a bus is a Poisson distribution. A Poisson distribution is suitable for modeling the number of events that occur in a fixed interval of time or space, given a known average rate of occurrence. In this case, the event is waiting for a bus, the interval is 10 minutes, and the average rate is 3 minutes. The Poisson distribution can capture the variability of the waiting time, which can range from 0 to 10 minutes, with different probabilities.

1: Poisson Distribution - Amazon SageMaker

2: Poisson Distribution - Wikipedia

 The simplest and fastest way to use the EFS dataset for SageMaker training is to run a SageMaker training job with an EFS file system as the data source. This option does not require any data copying or additional integration steps. SageMaker supports EFS as a data source for training jobs, and it can mount the EFS file system to the training container using the FileSystemConfig parameter. This way, the training script can access the data files as if they were on the local disk of the training instance. References:

Access Training Data - Amazon SageMaker

Mount an EFS file system to an Amazon SageMaker notebook (with lifecycle configurations) | AWS Machine Learning Blog

 Binary classification is a type of supervised learning problem where the goal is to predict a categorical label that has only two possible values, such as Yes or No, True or False, Positive or Negative. In this case, the label is whether a transaction is fraudulent or not, which is a binary outcome. Binary classification can be used to estimate the probability of an observation belonging to a certain class, such as the probability of a transaction being fraudulent. This can help the business to make decisions based on the risk level of each transaction. References:

Binary Classification - Amazon Machine Learning

AWS Certified Machine Learning - Specialty Sample Questions

To perform efficient exploratory data analysis (EDA) on a large dataset for anomaly detection, using the First K option in SageMaker Data Wrangler is an optimal choice. This option allows the data scientist to select the first K rows, limiting the data loaded into memory, which conserves compute resources.

Given that the First K option allows the data scientist to determine K based on domain knowledge, this approach provides a representative sample without requiring extensive compute resources. Other options like randomized sampling may not provide data samples that are as useful for initial analysis in a time-series or sequential dataset context.

A stratified k-fold cross-validation strategy is a technique that preserves the class distribution in each fold. This is important for imbalanced datasets, such as the one in the question, where the disease is seen in only 3% of the population. If a random k-fold cross-validation strategy is used, some folds may have no positive cases or very few, which would lead to poor estimates of the model performance. A stratified k-fold cross-validation strategy ensures that each fold has the same proportion of positive and negative cases as the whole dataset, which makes the evaluation more reliable and robust. A k-fold cross-validation strategy with k=5 and 3 repeats is also a possible option, but it is more computationally expensive and may not be necessary if the stratification is done properly. An 80/20 stratified split between training and validation is another option, but it uses less data for training and validation than k-fold cross-validation, which may result in higher variance and lower accuracy of the estimates. References:

AWS Machine Learning Specialty Certification Exam Guide

AWS Machine Learning Training: Model Evaluation

How to Fix k-Fold Cross-Validation for Imbalanced Classification

Principal component analysis (PCA) is a technique for reducing the dimensionality of datasets, increasing interpretability but at the same time minimizing information loss. It does so by creating new uncorrelated variables that successively maximize variance. PCA is useful when dealing with datasets that have a large number of correlated features. However, PCA is sensitive to the scale of the features, so it is important to standardize or normalize the data before applying PCA. Amazon SageMaker provides a built-in algorithm for PCA that can be used to transform the data into a lower-dimensional representation. Amazon SageMaker Data Wrangler is a tool that allows data scientists to visually explore, clean, and prepare data for machine learning. Data Wrangler provides various transformation steps that can be applied to the data, such as scaling, encoding, imputing, etc. Data Wrangler also integrates with SageMaker built-in algorithms, such as PCA, to enable feature engineering and dimensionality reduction. Therefore, option B is the correct answer, as it involves scaling the data with a Min Max Scaler transformation step, which rescales the data to a range of [0, 1], and then using the SageMaker built-in algorithm for PCA on the scaled dataset to transform the data. Option A is incorrect, as it does not involve scaling the data before applying PCA, which can affect the results of the dimensionality reduction. Option C is incorrect, as it involves removing the features that have the highest correlation, which can lead to information loss and reduce the performance of the regression model. Option D is incorrect, as it involves removing the features that have the lowest correlation, which can also lead to information loss and reduce the performance of the regression model. References:

Principal Component Analysis (PCA) - Amazon SageMaker

Scale data with a Min Max Scaler - Amazon SageMaker Data Wrangler

Use Amazon SageMaker built-in algorithms - Amazon SageMaker Data Wrangler

The best services to use for building a data ingestion solution for the company’s Amazon S3-based data lake are:

AWS Glue as the data catalog: AWS Glue is a fully managed extract, transform, and load (ETL) service that can discover, crawl, and catalog data from various sources and formats, and make it available for analysis. AWS Glue can also generate ETL code in Python or Scala to transform, enrich, and join data using AWS Glue Data Catalog as the metadata repository. AWS Glue Data Catalog is a central metadata store that integrates with Amazon Athena, Amazon EMR, and Amazon Redshift Spectrum, allowing users to create a unified view of their data across various sources and formats.

Amazon Kinesis Data Streams and Amazon Kinesis Data Analytics for real-time data insights: Amazon Kinesis Data Streams is a service that enables users to collect, process, and analyze real-time streaming data at any scale. Users can create data streams that can capture data from various sources, such as web and mobile applications, IoT devices, and social media platforms. Amazon Kinesis Data Analytics is a service that allows users to analyze streaming data using standard SQL queries or Apache Flink applications. Users can create real-time dashboards, metrics, and alerts based on the streaming data analysis results.

Amazon Kinesis Data Firehose for delivery to Amazon ES for clickstream analytics: Amazon Kinesis Data Firehose is a service that enables users to load streaming data into data lakes, data stores, and analytics services. Users can configure Kinesis Data Firehose to automatically deliver data to various destinations, such as Amazon S3, Amazon Redshift, Amazon OpenSearch Service, and third-party solutions. For clickstream analytics, users can use Kinesis Data Firehose to deliver data to Amazon OpenSearch Service, a fully managed service that offers search and analytics capabilities for log data. Users can use Amazon OpenSearch Service to perform interactive analysis and visualization of clickstream data using Kibana, an open-source tool that is integrated with Amazon OpenSearch Service.

Amazon EMR to generate personalized product recommendations: Amazon EMR is a service that enables users to run distributed data processing frameworks, such as Apache Spark, Apache Hadoop, and Apache Hive, on scalable clusters of EC2 instances. Users can use Amazon EMR to perform advanced analytics, such as machine learning, on large and complex datasets stored in Amazon S3 or other sources. For product recommendations, users can use Amazon EMR to run Spark MLlib, a library that provides scalable machine learning algorithms, such as collaborative filtering, to generate personalized recommendations based on user behavior and preferences.

AWS Glue - Fully Managed ETL Service

Amazon Kinesis - Data Streaming Service

Amazon OpenSearch Service - Managed OpenSearch Service

Amazon EMR - Managed Hadoop Framework

The correct solution for granting permissions for data preprocessing is to use the following steps:

Create an IAM role that has permissions to create Amazon SageMaker Processing jobs. Attach the role to the SageMaker notebook instance. This role allows the ML specialist to run Processing jobs from the notebook code1

Create an Amazon SageMaker Processing job with an IAM role that has read and write permissions to the relevant S3 bucket, and appropriate KMS and ECR permissions. This role allows the Processing job to access the data in the encrypted S3 bucket, decrypt it with the KMS CMK, and pull the container image from ECR23

The other options are incorrect because they either miss some permissions or use unnecessary steps. For example:

Option A uses a single IAM role for both the notebook instance and the Processing job. This role may have more permissions than necessary for the notebook instance, which violates the principle of least privilege4

Option C sets up both an S3 endpoint and a KMS endpoint in the default VPC. These endpoints are not required for the Processing job to access the data in the encrypted S3 bucket. They are only needed if the Processing job runs in network isolation mode, which is not specified in the question.

Option D uses the access key and secret key of the IAM user with appropriate KMS and ECR permissions. This is not a secure way to pass credentials to the Processing job. It also requires the ML specialist to manage the IAM user and the keys.

1: Create an Amazon SageMaker Notebook Instance - Amazon SageMaker

2: Create a Processing Job - Amazon SageMaker

3: Use AWS KMS–Managed Encryption Keys - Amazon Simple Storage Service

4: IAM Best Practices - AWS Identity and Access Management

Network Isolation - Amazon SageMaker

Understanding and Getting Your Security Credentials - AWS General Reference

For a model with low accuracy on both training and testing datasets, the following two strategies are effective:

Increase the number of passes and perform hyperparameter tuning: This approach allows the model to better learn from the existing data and improve performance through optimized hyperparameters.

Add domain-specific features and use more complex models: Adding relevant features that capture additional information from domain knowledge and using more complex model architectures can help the model capture patterns better, potentially improving accuracy.

Options B, D, and E would either reduce feature complexity or training data volume, which is less likely to improve performance when accuracy is low on both training and testing sets.

Amazon SageMaker Studio Data Wrangler is a visual data preparation tool that enables users to clean and normalize data without writing any code. Using Data Wrangler, the data scientist can easily import the time-series data from various sources, such as Amazon S3, Amazon Athena, or Amazon Redshift. Data Wrangler can automatically generate data insights and quality reports, which can help identify and fix missing values, outliers, and anomalies in the data. Data Wrangler also provides over 250 built-in transformations, such as resampling, interpolation, aggregation, and filtering, which can be applied to the data with a point-and-click interface. Data Wrangler can also export the prepared data to different destinations, such as Amazon S3, Amazon SageMaker Feature Store, or Amazon SageMaker Pipelines, for further modeling and analysis. Data Wrangler is integrated with Amazon SageMaker Studio, a web-based IDE for machine learning, which makes it easy to access and use the tool. Data Wrangler is a serverless and fully managed service, which means the data scientist does not need to provision, configure, or manage any infrastructure or clusters.

Option A is incorrect because Amazon EMR Serverless is a serverless option for running big data analytics applications using open-source frameworks, such as Apache Spark. However, using Amazon EMR Serverless would require the data scientist to write PySpark code to perform the data preparation tasks, such as resampling, imputation, and aggregation. This would require more implementation effort than using Data Wrangler, which provides a visual and code-free interface for data preparation.

Option B is incorrect because AWS Glue DataBrew is another visual data preparation tool that can be used to clean and normalize data without writing code. However, DataBrew does not support time-series data as a data type, and does not provide built-in transformations for resampling, interpolation, or aggregation of time-series data. Therefore, using DataBrew would not meet the requirements of the use case.

Option D is incorrect because using Amazon SageMaker Studio Notebook with Pandas would also require the data scientist to write Python code to perform the data preparation tasks. Pandas is a popular Python library for data analysis and manipulation, which supports time-series data and provides various methods for resampling, interpolation, and aggregation. However, using Pandas would require more implementation effort than using Data Wrangler, which provides a visual and code-free interface for data preparation.

1: Amazon SageMaker Data Wrangler documentation

2: Amazon EMR Serverless documentation

3: AWS Glue DataBrew documentation

4: Pandas documentation

K-means clustering is a machine learning technique that can be used to classify unlabeled data into different groups based on their similarity. It is an unsupervised learning method, which means it does not require any prior knowledge or labels for the data. K-means clustering works by randomly assigning data points to a number of clusters, then iteratively updating the cluster centers and reassigning the data points until the clusters are stable. The result is a partition of the data into distinct and homogeneous groups. K-means clustering can be useful for exploratory data analysis, data compression, anomaly detection, and feature extraction. References:

K-Means Clustering: A tutorial on how to use K-means clustering with Amazon SageMaker.

Unsupervised Learning: A video that explains the concept and applications of unsupervised learning.

To evaluate the contribution of each feature before model training, feature importance scores are a primary tool. In Amazon SageMaker Data Wrangler, a Quick Model can be used to rapidly generate a model on the dataset and return feature importance scores. These scores indicate how much each feature influences the target variable.

“Use Quick Model to generate feature importance scores quickly to help understand which features contribute the most to predictions. These scores range from 0 to 1, with higher values indicating higher contribution.”

This allows ML engineers to assess feature relevance without full-scale model development, thus saving time and effort in the feature selection phase of the pipeline.

The best options to use an Inception neural network architecture instead of a ResNet architecture for image classification in Amazon SageMaker are:

Bundle a Docker container with TensorFlow Estimator loaded with an Inception network and use this for model training. This option allows users to customize the training environment and use any TensorFlow model they want. Users can create a Docker image that contains the TensorFlow Estimator API and the Inception model from the TensorFlow Hub, and push it to Amazon ECR. Then, users can use the SageMaker Estimator class to train the model using the custom Docker image and the training data from Amazon S3.

Use custom code in Amazon SageMaker with TensorFlow Estimator to load the model with an Inception network and use this for model training. This option allows users to use the built-in TensorFlow container provided by SageMaker and write custom code to load and train the Inception model. Users can use the TensorFlow Estimator class to specify the custom code and the training data from Amazon S3. The custom code can use the TensorFlow Hub module to load the Inception model and fine-tune it on the training data.

The other options are not feasible for this scenario because:

Customize the built-in image classification algorithm to use Inception and use this for model training. This option is not possible because the built-in image classification algorithm in SageMaker does not support customizing the neural network architecture. The built-in algorithm only supports ResNet models with different depths and widths.

Create a support case with the SageMaker team to change the default image classification algorithm to Inception. This option is not realistic because the SageMaker team does not provide such a service. Users cannot request the SageMaker team to change the default algorithm or add new algorithms to the built-in ones.

Download and apt-get install the inception network code into an Amazon EC2 instance and use this instance as a Jupyter notebook in Amazon SageMaker. This option is not advisable because it does not leverage the benefits of SageMaker, such as managed training and deployment, distributed training, and automatic model tuning. Users would have to manually install and configure the Inception network code and the TensorFlow framework on the EC2 instance, and run the training and inference code on the same instance, which may not be optimal for performance and scalability.

Use Your Own Algorithms or Models with Amazon SageMaker

Use the SageMaker TensorFlow Serving Container

TensorFlow Hub

 Pushing the data from Microsoft SQL Server to Amazon S3 using an AWS Data Pipeline and providing the S3 location within the notebook is the best approach for training a model using the data stored in Amazon RDS. This is because Amazon SageMaker can directly access data from Amazon S3 and train models on it. AWS Data Pipeline is a service that can automate the movement and transformation of data between different AWS services. It can also use Amazon RDS as a data source and Amazon S3 as a data destination. This way, the data can be transferred efficiently and securely without writing any code within the notebook. References:

Amazon SageMaker

AWS Data Pipeline

The best way to meet the requirements is to use a principal component analysis (PCA) model, which is a technique that reduces the dimensionality of the dataset by transforming the original variables into a smaller set of new variables, called principal components, that capture most of the variance and information in the data1. This technique has the following advantages:

It can increase the accuracy of the model by removing noise, redundancy, and multicollinearity from the data, and by enhancing the interpretability and generalization of the model23.

It can decrease the processing time of the model by reducing the number of features and the computational complexity of the model, and by improving the convergence and stability of the model45.

It is suitable for numeric variables, as it relies on the covariance or correlation matrix of the data, and it can handle a large quantity of variables, as it can extract the most relevant ones16.

The other options are not effective or appropriate, because they have the following drawbacks:

A: Creating new features and interaction variables can increase the accuracy of the model by capturing more complex and nonlinear relationships in the data, but it can also increase the processing time of the model by adding more features and increasing the computational complexity of the model7. Moreover, it can introduce more noise, redundancy, and multicollinearity in the data, which can degrade the performance and interpretability of the model8.

C: Applying normalization on the feature set can increase the accuracy of the model by scaling the features to a common range and avoiding the dominance of some features over others, but it can also decrease the processing time of the model by reducing the numerical instability and improving the convergence of the model . However, normalization alone is not enough to address the high dimensionality and high latency issues of the dataset, as it does not reduce the number of features or the variance in the data.

D: Using a multiple correspondence analysis (MCA) model is not suitable for numeric variables, as it is a technique that reduces the dimensionality of the dataset by transforming the original categorical variables into a smaller set of new variables, called factors, that capture most of the inertia and information in the data. MCA is similar to PCA, but it is designed for nominal or ordinal variables, not for continuous or interval variables.

1: Principal Component Analysis - Amazon SageMaker

2: How to Use PCA for Data Visualization and Improved Performance in Machine Learning | by Pratik Shukla | Towards Data Science

3: Principal Component Analysis (PCA) for Feature Selection and some of its Pitfalls | by Nagesh Singh Chauhan | Towards Data Science

4: How to Reduce Dimensionality with PCA and Train a Support Vector Machine in Python | by James Briggs | Towards Data Science

5: Dimensionality Reduction and Its Applications | by Aniruddha Bhandari | Towards Data Science

6: Principal Component Analysis (PCA) in Python | by Susan Li | Towards Data Science

7: Feature Engineering for Machine Learning | by Dipanjan (DJ) Sarkar | Towards Data Science

8: Feature Engineering — How to Engineer Features and How to Get Good at It | by Parul Pandey | Towards Data Science

[Feature Scaling for Machine Learning: Understanding the Difference Between Normalization vs. Standardization | by Benjamin Obi Tayo Ph.D. | Towards Data Science]

[Why, How and When to Scale your Features | by George Seif | Towards Data Science]

[Normalization vs Dimensionality Reduction | by Saurabh Annadate | Towards Data Science]

[Multiple Correspondence Analysis - Amazon SageMaker]

[Multiple Correspondence Analysis (MCA) | by Raul Eulogio | Towards Data Science]

In this scenario, the company requires that access to the Amazon SageMaker notebook remain within the AWS internal network, avoiding the public internet. By creating a VPC interface endpoint for SageMaker, the company can ensure that traffic to the SageMaker notebook remains internal to the VPC and is accessible over a private connection. The VPC interface endpoint allows private network access to AWS services, and it operates over AWS’s internal network, respecting the security and connectivity policies the company requires.

This solution requires minimal development effort compared to options involving bastion hosts or NAT gateways, as it directly provides private network access to the SageMaker notebook.

The task involves periodic inference (hourly batches of telemetry data) that can be processed in bulk, with the total processing time under 5 minutes. The most cost-effective solution for this batch prediction scenario is SageMaker Batch Transform, as it avoids the need for always-on endpoints.

“Use Batch Transform when you: … Don’t need a persistent endpoint and want to process batches of data at once. This is typically the most cost-efficient way to process large volumes of input data periodically.”

Since the use case doesn't require low latency or asynchronous real-time response, and processing takes only minutes, Batch Transform is the optimal fit.

Amazon SageMaker script mode is a feature that allows users to use training scripts similar to those they would use outside SageMaker with SageMaker’s prebuilt containers for various frameworks such as TensorFlow. Script mode supports reading data from Amazon S3 buckets without requiring any changes to the training script. Therefore, option B is the best method of providing training data to Amazon SageMaker that would meet the business requirements with the least development overhead.

Option A is incorrect because using a local path of the data would not be scalable or reliable, as it would depend on the availability and capacity of the local storage. Moreover, using a local path of the data would not leverage the benefits of Amazon S3, such as durability, security, and performance. Option C is incorrect because rewriting the train.py script to convert TFRecords to protobuf would require additional development effort and complexity, as well as introduce potential errors and inconsistencies in the data format. Option D is incorrect because preparing the data in the format accepted by Amazon SageMaker would also require additional development effort and complexity, as well as involve using additional services such as AWS Glue or AWS Lambda, which would increase the cost and maintenance of the solution.

Bring your own model with Amazon SageMaker script mode

GitHub - aws-samples/amazon-sagemaker-script-mode

Deep Dive on TensorFlow training with Amazon SageMaker and Amazon S3

amazon-sagemaker-script-mode/generate_cifar10_tfrecords.py at master

 The F1 score is a measure of the harmonic mean of precision and recall, which are both important for fraud detection. Precision is the ratio of true positives to all predicted positives, and recall is the ratio of true positives to all actual positives. A high F1 score indicates that the classifier can correctly identify fraudulent transactions and avoid false negatives. The true positive rate is another name for recall, and it measures the proportion of fraudulent transactions that are correctly detected by the classifier. A high true positive rate means that the classifier can capture as many fraudulent transactions as possible.

Fraud Detection Using Machine Learning | Implementations | AWS Solutions

Detect fraudulent transactions using machine learning with Amazon SageMaker | AWS Machine Learning Blog

1. Introduction — Reproducible Machine Learning for Credit Card Fraud Detection

SageMaker Data Wrangler is a feature of SageMaker Studio that provides an end-to-end solution for importing, preparing, transforming, featurizing, and analyzing data. Data Wrangler includes built-in analyses that help generate visualizations and data insights in a few clicks. One of the built-in analyses is the Quick Model visualization, which can be used to quickly evaluate the data and produce importance scores for each feature. A feature importance score indicates how useful a feature is at predicting a target label. The feature importance score is between [0, 1] and a higher number indicates that the feature is more important to the whole dataset. The Quick Model visualization uses a random forest model to calculate the feature importance for each feature using the Gini importance method. This method measures the total reduction in node impurity (a measure of how well a node separates the classes) that is attributed to splitting on a particular feature. The ML developer can use the Quick Model visualization to obtain the importance scores for each feature of the dataset and use them to feature engineer the dataset. This solution requires the least development effort compared to the other options.

The Data Scientist should increase the XGBoost scale_pos_weight parameter to adjust the balance of positive and negative weights and change the XGBoost eval_metric parameter to optimize based on Area Under the ROC Curve (AUC). This will help reduce the number of false negative predictions by the model.

The scale_pos_weight parameter controls the balance of positive and negative weights in the XGBoost algorithm. It is useful for imbalanced classification problems, such as fraud detection, where the number of positive examples (fraudulent transactions) is much smaller than the number of negative examples (non-fraudulent transactions). By increasing the scale_pos_weight parameter, the Data Scientist can assign more weight to the positive class and make the model more sensitive to detecting fraudulent transactions.

The eval_metric parameter specifies the metric that is used to measure the performance of the model during training and validation. The default metric for binary classification problems is the error rate, which is the fraction of incorrect predictions. However, the error rate is not a good metric for imbalanced classification problems, because it does not take into account the cost of different types of errors. For example, in fraud detection, a false negative (failing to detect a fraudulent transaction) is more costly than a false positive (flagging a non-fraudulent transaction as fraudulent). Therefore, the Data Scientist should use a metric that reflects the trade-off between the true positive rate (TPR) and the false positive rate (FPR), such as the Area Under the ROC Curve (AUC). The AUC is a measure of how well the model can distinguish between the positive and negative classes, regardless of the classification threshold. A higher AUC means that the model can achieve a higher TPR with a lower FPR, which is desirable for fraud detection.

XGBoost Parameters - Amazon Machine Learning

Using XGBoost with Amazon SageMaker - AWS Machine Learning Blog

A logarithmic transformation is a suitable data transformation for a linear regression model when the data has a skewed distribution, such as when the mode is lower than the median and the median is lower than the mean. A logarithmic transformation can reduce the skewness and make the data more symmetric and normally distributed, which are desirable properties for linear regression. A logarithmic transformation can also reduce the effect of outliers and heteroscedasticity (unequal variance) in the data. An exponential transformation would have the opposite effect of increasing the skewness and making the data more asymmetric. A polynomial transformation may not be able to capture the nonlinearity in the data and may introduce multicollinearity among the transformed variables. A sinusoidal transformation is not appropriate for data that does not have a periodic pattern.

Data Transformation - Scaler Topics

Linear Regression - GeeksforGeeks

Linear Regression - Scribbr

The input mode for the training job determines how the training data is transferred from Amazon S3 to the SageMaker instance. There are two input modes: File and Pipe. File mode copies the entire training dataset from S3 to the local file system of the instance before starting the training job. This can cause a long delay before the training job launches, especially if the dataset is large. Pipe mode streams the data from S3 to the instance as the training job runs. This can reduce the startup time and improve the I/O throughput, as the data is read in smaller batches. Therefore, to address the performance issues with the current solution, the Specialist should ensure that the input mode for the training job is set to Pipe. This can be done by using the SageMaker Python SDK and setting the input_mode parameter to Pipe when creating the estimator or the fit method12. Alternatively, this can be done by using the AWS CLI and setting the InputMode parameter to Pipe when creating the training job3.

Access Training Data - Amazon SageMaker

Choosing Data Input Mode Using the SageMaker Python SDK - Amazon SageMaker

CreateTrainingJob - Amazon SageMaker Service

Principal component analysis (PCA) is an unsupervised machine learning algorithm that attempts to reduce the dimensionality (number of features) within a dataset while still retaining as much information as possible. This is done by finding a new set of features called components, which are composites of the original features that are uncorrelated with one another. They are also constrained so that the first component accounts for the largest possible variability in the data, the second component the second most variability, and so on. By using PCA, the impact of having a large number of features that are highly correlated with each other can be reduced, as the new feature space will have fewer dimensions and less redundancy. This can make the linear models more stable and less prone to overfitting. References:

Principal Component Analysis (PCA) Algorithm - Amazon SageMaker

Perform a large-scale principal component analysis faster using Amazon SageMaker | AWS Machine Learning Blog

Machine Learning- Prinicipal Component Analysis | i2tutorials

A sequential split is a method of splitting data into training and evaluation datasets while preserving the order of the data records. This method is useful when the data has a temporal or sequential structure, and the order of the data matters for the prediction task. For example, if the data contains sales data for different months or years, and the goal is to predict the sales for the next month or year, a sequential split can ensure that the training data comes from the earlier period and the evaluation data comes from the later period. This can help avoid data leakage, which occurs when the training data contains information from the future that is not available at the time of prediction. A sequential split can also help evaluate the model performance on the most recent data, which may be more relevant and representative of the future data.

In this question, Example Corp has sequential sales data from the past 15 years and wants to use Amazon ML to predict the sales for this year’s upcoming annual sale event. A sequential split is the most appropriate method for splitting the data, as it can preserve the order of the data and prevent data leakage. For example, Example Corp can use the data from the first 14 years as the training dataset, and the data from the last year as the evaluation dataset. This way, the model can learn from the historical data and be tested on the most recent data.

Amazon ML provides an option to split the data sequentially when creating the training and evaluation datasources. To use this option, Example Corp can specify the percentage of the data to use for training and evaluation, and Amazon ML will use the first part of the data for training and the remaining part of the data for evaluation. For more information, see Splitting Your Data - Amazon Machine Learning.

Amazon Athena is an interactive query service that allows you to analyze data stored in Amazon S3 using standard SQL. Athena is serverless, so you only pay for the queries that you run and there is no infrastructure to manage.

To optimize the query performance of Athena, one of the best practices is to convert the data into a columnar format, such as Apache Parquet or Apache ORC. Columnar formats store data by columns rather than by rows, which allows Athena to scan only the columns that are relevant to the query, reducing the amount of data read and improving the query speed. Columnar formats also support compression and encoding schemes that can reduce the storage space and the data scanned per query, further enhancing the performance and reducing the cost.

In contrast, plaintext CSV files store data by rows, which means that Athena has to scan the entire row even if only a few columns are needed for the query. This increases the amount of data read and the query latency. Moreover, plaintext CSV files do not support compression or encoding, which means that they take up more storage space and incur higher query costs.

Therefore, the Machine Learning Specialist should transform the dataset to Apache Parquet format to minimize query runtime.

Top 10 Performance Tuning Tips for Amazon Athena

Columnar Storage Formats

Using compressions will reduce the amount of data scanned by Amazon Athena, and also reduce your S3 bucket storage. It’s a Win-Win for your AWS bill. Supported formats: GZIP, LZO, SNAPPY (Parquet) and ZLIB.

 A receiver operating characteristic (ROC) curve is a model evaluation technique that can be used to understand how different classification thresholds will impact the model’s performance. A ROC curve plots the true positive rate (TPR) against the false positive rate (FPR) for various values of the classification threshold. The TPR, also known as sensitivity or recall, is the proportion of positive instances that are correctly classified as positive. The FPR, also known as the fall-out, is the proportion of negative instances that are incorrectly classified as positive. A ROC curve can show the trade-off between the TPR and the FPR for different thresholds, and help the Machine Learning Specialist to select the optimal threshold that maximizes the TPR and minimizes the FPR. A ROC curve can also be used to compare the performance of different models by calculating the area under the curve (AUC), which is a measure of how well the model can distinguish between the positive and negative classes. A higher AUC indicates a better model

The best architecture to scale the solution at the lowest cost is to implement the solution using AWS Deep Learning Containers and run the container as a job using AWS Batch on a GPU-compatible Spot Instance. This option has the following advantages:

AWS Deep Learning Containers: These are Docker images that are pre-installed and optimized with popular deep learning frameworks such as TensorFlow, PyTorch, and MXNet. They can be easily deployed on Amazon EC2, Amazon ECS, Amazon EKS, and AWS Fargate. They can also be integrated with AWS Batch to run containerized batch jobs. Using AWS Deep Learning Containers can simplify the setup and configuration of the deep learning environment and reduce the complexity of the resource management.

AWS Batch: This is a fully managed service that enables you to run batch computing workloads on AWS. You can define compute environments, job queues, and job definitions to run your batch jobs. You can also use AWS Batch to automatically provision compute resources based on the requirements of the batch jobs. You can specify the type and quantity of the compute resources, such as GPU instances, and the maximum price you are willing to pay for them. You can also use AWS Batch to monitor the status and progress of your batch jobs and handle any failures or interruptions.

GPU-compatible Spot Instance: This is an Amazon EC2 instance that uses a spare compute capacity that is available at a lower price than the On-Demand price. You can use Spot Instances to run your deep learning training jobs at a lower cost, as long as you are flexible about when your instances run and how long they run. You can also use Spot Instances with AWS Batch to automatically launch and terminate instances based on the availability and price of the Spot capacity. You can also use Spot Instances with Amazon EBS volumes to store your datasets, checkpoints, and logs, and attach them to your instances when they are launched. This way, you can preserve your data and resume your training even if your instances are interrupted.

AWS Deep Learning Containers

AWS Batch

Amazon EC2 Spot Instances

Using Amazon EBS Volumes with Amazon EC2 Spot Instances

The Machine Learning team wants to use its own training algorithm on Amazon SageMaker, and submit both its own algorithm code and algorithm-specific parameters. The best combination of services to build a custom algorithm in Amazon SageMaker are Amazon ECR and Amazon S3.

Amazon ECR is a fully managed container registry service that allows you to store, manage, and deploy Docker container images. You can use Amazon ECR to create a Docker image that contains your training algorithm code and any dependencies or libraries that it requires. You can also use Amazon ECR to push, pull, and manage your Docker images securely and reliably.

Amazon S3 is a durable, scalable, and secure object storage service that can store any amount and type of data. You can use Amazon S3 to store your training data, model artifacts, and algorithm-specific parameters. You can also use Amazon S3 to access your data and parameters from your training algorithm code, and to write your model output to a specified location.

Therefore, the Machine Learning team can use the following steps to build a custom algorithm in Amazon SageMaker:

Write the training algorithm code in Python, using the Amazon SageMaker Python SDK or the Amazon SageMaker Containers library to interact with the Amazon SageMaker service. The code should be able to read the input data and parameters from Amazon S3, and write the model output to Amazon S3.

Create a Dockerfile that defines the base image, the dependencies, the environment variables, and the commands to run the training algorithm code. The Dockerfile should also expose the ports that Amazon SageMaker uses to communicate with the container.

Build the Docker image using the Dockerfile, and tag it with a meaningful name and version.

Push the Docker image to Amazon ECR, and note the registry path of the image.

Upload the training data, model artifacts, and algorithm-specific parameters to Amazon S3, and note the S3 URIs of the objects.

Create an Amazon SageMaker training job, using the Amazon SageMaker Python SDK or the AWS CLI. Specify the registry path of the Docker image, the S3 URIs of the input and output data, the algorithm-specific parameters, and other configuration options, such as the instance type, the number of instances, the IAM role, and the hyperparameters.

Monitor the status and logs of the training job, and retrieve the model output from Amazon S3.

Use Your Own Training Algorithms

Amazon ECR - Amazon Web Services

Amazon S3 - Amazon Web Services

The metrics that will indicate an ML solution that will provide the greatest probability of detecting an abnormality are precision and recall. Precision is the ratio of true positives (TP) to the total number of predicted positives (TP + FP), where FP is false positives. Recall is the ratio of true positives (TP) to the total number of actual positives (TP + FN), where FN is false negatives. A high precision means that the ML solution has a low rate of false alarms, while a high recall means that the ML solution has a high rate of true detections. For the chemical company, the goal is to avoid process abnormalities, which are marked as 1 in the labels. Therefore, the company needs an ML solution that has a high recall for the positive class, meaning that it can detect most of the abnormalities and minimize the false negatives. Among the four options, option B has the highest recall for the positive class, which is 0.98. This means that the ML solution can detect 98% of the abnormalities and miss only 2%. Option B also has a reasonable precision for the positive class, which is 0.61. This means that the ML solution has a false alarm rate of 39%, which may be acceptable for the company, depending on the cost and benefit analysis. The other options have lower recall for the positive class, which means that they have higher false negative rates, which can be more detrimental for the company than false positive rates.

1: AWS Certified Machine Learning - Specialty Exam Guide

2: AWS Training - Machine Learning on AWS

3: AWS Whitepaper - An Overview of Machine Learning on AWS

4: Precision and recall

The best solution for this scenario is to use a multi-model endpoint in Amazon SageMaker, which allows hosting multiple models on the same endpoint and invoking them dynamically at runtime. This way, the company can reduce the operational overhead of managing multiple EC2 instances and model servers, and leverage the scalability, security, and performance of SageMaker hosting services. By using a multi-model endpoint, the company can also save on hosting costs by improving endpoint utilization and paying only for the models that are loaded in memory and the API calls that are made. To use a multi-model endpoint, the company needs to prepare a Docker container based on the open-source multi-model server, which is a framework-agnostic library that supports loading and serving multiple models from Amazon S3. The company can then create a multi-model endpoint in SageMaker, pointing to the S3 bucket containing all the models, and invoke the endpoint from the web client at runtime, specifying the TargetModel parameter according to the city of each request. This solution also enables the company to add or remove models from the S3 bucket without redeploying the endpoint, and to use different versions of the same model for different cities if needed. References:

Use Docker containers to build models

Host multiple models in one container behind one endpoint

Multi-model endpoints using Scikit Learn

Multi-model endpoints using XGBoost

Option A is correct because reducing the number of features with the SageMaker PCA algorithm can help remove noise and redundancy from the data, and improve the model’s performance. PCA is a dimensionality reduction technique that transforms the original features into a smaller set of linearly uncorrelated features called principal components. The SageMaker linear learner algorithm supports PCA as a built-in feature transformation option.

Option E is correct because using the SageMaker k-NN algorithm with a dimension reduction target of less than 1,000 can help the model learn from the similarity of the data points, and improve the model’s performance. k-NN is a non-parametric algorithm that classifies an input based on the majority vote of its k nearest neighbors in the feature space. The SageMaker k-NN algorithm supports dimension reduction as a built-in feature transformation option.

Option B is incorrect because using the scikit-learn MDS algorithm to reduce the number of features is not a feasible option, as MDS is a computationally expensive technique that does not scale well to large datasets. MDS is a dimensionality reduction technique that tries to preserve the pairwise distances between the original data points in a lower-dimensional space.

Option C is incorrect because setting the predictor type to regressor would change the model’s objective from classification to regression, which is not suitable for the given problem. A regressor model would output a continuous value instead of a binary label for each phone.

Option D is incorrect because using the SageMaker k-means algorithm with k of less than 1,000 would not help the model classify the phones, as k-means is a clustering algorithm that groups the data points into k clusters based on their similarity, without using any labels. A clustering model would not output a binary label for each phone.

Amazon SageMaker Linear Learner Algorithm

Amazon SageMaker K-Nearest Neighbors (k-NN) Algorithm

[Principal Component Analysis - Scikit-learn]

[Multidimensional Scaling - Scikit-learn]

The combination of steps that will meet the requirements are to create an IAM role in the development account that the integration account and production account can assume, attach IAM policies to the role that allow access to the feature repository and the S3 buckets, and share the feature repository that is associated with the S3 buckets from the development account to the integration account and the production account by using AWS Resource Access Manager (AWS RAM). This approach will enable cross-account access and sharing of the features stored in Amazon SageMaker Feature Store and Amazon S3.

Amazon SageMaker Feature Store is a fully managed, purpose-built repository to store, update, search, and share curated data used in training and prediction workflows. The service provides feature management capabilities such as enabling easy feature reuse, low latency serving, time travel, and ensuring consistency between features used in training and inference workflows. A feature group is a logical grouping of ML features whose organization and structure is defined by a feature group schema. A feature group schema consists of a list of feature definitions, each of which specifies the name, type, and metadata of a feature. Amazon SageMaker Feature Store stores the features in both an online store and an offline store. The online store is a low-latency, high-throughput store that is optimized for real-time inference. The offline store is a historical store that is backed by an Amazon S3 bucket and is optimized for batch processing and model training1.

AWS Identity and Access Management (IAM) is a web service that helps you securely control access to AWS resources for your users. You use IAM to control who can use your AWS resources (authentication) and what resources they can use and in what ways (authorization). An IAM role is an IAM identity that you can create in your account that has specific permissions. You can use an IAM role to delegate access to users, applications, or services that don’t normally have access to your AWS resources. For example, you can create an IAM role in your development account that allows the integration account and the production account to assume the role and access the resources in the development account. You can attach IAM policies to the role that specify the permissions for the feature repository and the S3 buckets. You can also use IAM conditions to restrict the access based on the source account, IP address, or other factors2.

AWS Resource Access Manager (AWS RAM) is a service that enables you to easily and securely share AWS resources with any AWS account or within your AWS Organization. You can share AWS resources that you own with other accounts using resource shares. A resource share is an entity that defines the resources that you want to share, and the principals that you want to share with. For example, you can share the feature repository that is associated with the S3 buckets from the development account to the integration account and the production account by creating a resource share in AWS RAM. You can specify the feature group ARN and the S3 bucket ARN as the resources, and the integration account ID and the production account ID as the principals. You can also use IAM policies to further control the access to the shared resources3.

The other options are either incorrect or unnecessary. Using AWS Security Token Service (AWS STS) from the integration account and the production account to retrieve credentials for the development account is not required, as the IAM role in the development account can provide temporary security credentials for the cross-account access. Setting up S3 replication between the development S3 buckets and the integration and production S3 buckets would introduce redundancy and inconsistency, as the S3 buckets are already shared through AWS RAM. Creating an AWS PrivateLink endpoint in the development account for SageMaker is not relevant, as it is used to securely connect to SageMaker services from a VPC, not from another account.

1: Amazon SageMaker Feature Store – Amazon Web Services

2: What Is IAM? - AWS Identity and Access Management

3: What Is AWS Resource Access Manager? - AWS Resource Access Manager

The Data Science team is facing a problem of imbalanced data, where the positive class (paid users) is much less frequent than the negative class (non-paid users). This can cause the random forest model to be biased towards the majority class and have poor performance on the minority class. To mitigate this issue, the Data Science team can try the following approaches:

C. Generate more positive samples by duplicating the positive samples and adding a small amount of noise to the duplicated data. This is a technique called data augmentation, which can help increase the size and diversity of the training data for the minority class. This can help the random forest model learn more features and patterns from the positive class and reduce the imbalance ratio.

D. Change the cost function so that false negatives have a higher impact on the cost value than false positives. This is a technique called cost-sensitive learning, which can assign different weights or costs to different classes or errors. By assigning a higher cost to false negatives (predicting non-paid when the user is actually paid), the random forest model can be more sensitive to the minority class and try to minimize the misclassification of the positive class.

Bagging and Random Forest for Imbalanced Classification

Surviving in a Random Forest with Imbalanced Datasets

machine learning - random forest for imbalanced data? - Cross Validated

Biased Random Forest For Dealing With the Class Imbalance Problem

SageMaker Automatic Model Tuning (AMT) is a fully managed hyperparameter optimization feature that finds the best values for model parameters like eta, alpha, min_child_weight, and max_depth.

“Use SageMaker automatic model tuning to find the best version of a model by running many training jobs on your dataset using the ranges of hyperparameters that you specify.”

It supports built-in algorithms like XGBoost, and can optimize for evaluation metrics like AUC, making it the least operational overhead solution for this task.

The best storage scheme for this scenario is to store datasets as files in Amazon S3. Amazon S3 is a scalable, cost-effective, and durable object storage service that can store any amount and type of data. Amazon S3 also supports querying data using SQL with Amazon Athena, a serverless interactive query service that can analyze data directly in S3. This way, the Data Science team can easily explore and analyze their datasets without having to load them into a database or a compute instance.

The other options are not as suitable for this scenario because:

Storing datasets as files in an Amazon EBS volume attached to an Amazon EC2 instance would limit the scalability and availability of the data, as EBS volumes are only accessible within a single availability zone and have a maximum size of 16 TiB. Also, EBS volumes are more expensive than S3 buckets and require provisioning and managing EC2 instances.

Storing datasets as tables in a multi-node Amazon Redshift cluster would incur higher costs and complexity than using S3 and Athena. Amazon Redshift is a data warehouse service that is optimized for analytical queries over structured or semi-structured data. However, it requires setting up and maintaining a cluster of nodes, loading data into tables, and choosing the right distribution and sort keys for optimal performance. Moreover, Amazon Redshift charges for both storage and compute, while S3 and Athena only charge for the amount of data stored and scanned, respectively.

Storing datasets as global tables in Amazon DynamoDB would not be feasible for large amounts of data, as DynamoDB is a key-value and document database service that is designed for fast and consistent performance at any scale. However, DynamoDB has a limit of 400 KB per item and 25 GB per partition key value, which may not be enough for storing large datasets. Also, DynamoDB does not support SQL queries natively, and would require using a service like Amazon EMR or AWS Glue to run SQL queries over DynamoDB data.

Amazon S3 - Cloud Object Storage

Amazon Athena – Interactive SQL Queries for Data in Amazon S3

Amazon EBS - Amazon Elastic Block Store (EBS)

Amazon Redshift – Data Warehouse Solution - AWS

Amazon DynamoDB – NoSQL Cloud Database Service

To develop a daily ETL workflow containing multiple ETL jobs that can start as soon as data is uploaded to Amazon S3, the best configuration is to use AWS Lambda to trigger an AWS Step Functions workflow to wait for dataset uploads to complete in Amazon S3. Use AWS Glue to join the datasets. Use an Amazon CloudWatch alarm to send an SNS notification to the Administrator in the case of a failure.

AWS Lambda is a serverless compute service that lets you run code without provisioning or managing servers. You can use Lambda to create functions that respond to events such as data uploads to Amazon S3. You can also use Lambda to invoke other AWS services such as AWS Step Functions and AWS Glue.

AWS Step Functions is a service that lets you coordinate multiple AWS services into serverless workflows. You can use Step Functions to create a state machine that defines the sequence and logic of your ETL workflow. You can also use Step Functions to handle errors and retries, and to monitor the execution status of your workflow.

AWS Glue is a serverless data integration service that makes it easy to discover, prepare, and combine data for analytics. You can use Glue to create and run ETL jobs that can join data from multiple sources in Amazon S3. You can also use Glue to catalog your data and make it searchable and queryable.

Amazon CloudWatch is a service that monitors your AWS resources and applications. You can use CloudWatch to create alarms that trigger actions when a metric or a log event meets a specified threshold. You can also use CloudWatch to send notifications to Amazon Simple Notification Service (SNS) topics, which can then deliver the notifications to subscribers such as email addresses or phone numbers.

Therefore, by using these services together, you can achieve the following benefits:

You can start the ETL workflow as soon as data is uploaded to Amazon S3 by using Lambda functions to trigger Step Functions workflows.

You can wait for all the datasets to be available in Amazon S3 by using Step Functions to poll the S3 buckets and check the data completeness.

You can join the datasets with terabyte-sized datasets in Amazon S3 by using Glue ETL jobs that can scale and parallelize the data processing.

You can store the results of joining datasets in Amazon S3 by using Glue ETL jobs to write the output to S3 buckets.

You can send a notification to the Administrator if one of the jobs fails by using CloudWatch alarms to monitor the Step Functions or Glue metrics and send SNS notifications in case of a failure.

Regularization is a technique that can be used to prevent overfitting and improve model performance on unseen data. Overfitting occurs when the model learns the training data too well and fails to generalize to new and unseen data. This can be seen in the question, where the validation accuracy is lower than the training accuracy, and both are lower than the human-level performance. Regularization is a way of adding some constraints or penalties to the model to reduce its complexity and prevent it from memorizing the training data. Some common forms of regularization for image classification are:

Weight decay: Adding a term to the loss function that penalizes large weights in the model. This can help reduce the variance and noise in the model and make it more robust to small changes in the input.

Dropout: Randomly dropping out some units or connections in the model during training. This can help reduce the co-dependency among the units and make the model more resilient to missing or corrupted features.

Data augmentation: Artificially increasing the size and diversity of the training data by applying random transformations, such as cropping, flipping, rotating, scaling, etc. This can help the model learn more invariant and generalizable features and reduce the risk of overfitting to specific patterns in the training data.

The other options are not likely to fix the issue of overfitting, and may even worsen it:

A longer training time: This can lead to more overfitting, as the model will have more chances to fit the noise and details in the training data that are not relevant for the validation data.

Making the network larger: This can increase the model capacity and complexity, which can also lead to more overfitting, as the model will have more parameters to learn and adjust to the training data.

Using a different optimizer: This can affect the speed and stability of the training process, but not necessarily the generalization ability of the model. The choice of optimizer depends on the characteristics of the data and the model, and there is no guarantee that a different optimizer will prevent overfitting.

Regularization (machine learning)

Image Classification: Regularization

How to Reduce Overfitting With Dropout Regularization in Keras

 The solution A will meet the requirements with the least amount of effort from the internal team because it uses Amazon SageMaker Ground Truth and Amazon Rekognition Custom Labels, which are fully managed services that can provide the desired functionality. The solution A involves the following steps:

Set up a private workforce that consists of the internal team. Use the private workforce and the SageMaker Ground Truth active learning feature to label the data. Amazon SageMaker Ground Truth is a service that can create high-quality training datasets for machine learning by using human labelers. A private workforce is a group of labelers that the company can manage and control. The internal team can use the private workforce to label the satellite images as having solar panels or not. The SageMaker Ground Truth active learning feature can reduce the labeling effort by using a machine learning model to automatically label the easy examples and only send the difficult ones to the human labelers1.

Use Amazon Rekognition Custom Labels for model training and hosting. Amazon Rekognition Custom Labels is a service that can train and deploy custom machine learning models for image analysis. Amazon Rekognition Custom Labels can use the labeled data from SageMaker Ground Truth to train a model that can detect solar panels in satellite images. Amazon Rekognition Custom Labels can also host the model and provide an API endpoint for inference2.

The other options are not suitable because:

Option B: Setting up a private workforce that consists of the internal team, using the private workforce to label the data, and using Amazon Rekognition Custom Labels for model training and hosting will incur more effort from the internal team than using SageMaker Ground Truth active learning feature. The internal team will have to label all the images manually, without the assistance of the machine learning model that can automate some of the labeling tasks1.

Option C: Setting up a private workforce that consists of the internal team, using the private workforce and the SageMaker Ground Truth active learning feature to label the data, using the SageMaker Object Detection algorithm to train a model, and using SageMaker batch transform for inference will incur more operational overhead than using Amazon Rekognition Custom Labels. The company will have to manage the SageMaker training job, the model artifact, and the batch transform job. Moreover, SageMaker batch transform is not suitable for real-time inference, as it processes the data in batches and stores the results in Amazon S33.

Option D: Setting up a public workforce, using the public workforce to label the data, using the SageMaker Object Detection algorithm to train a model, and using SageMaker batch transform for inference will incur more operational overhead and cost than using a private workforce and Amazon Rekognition Custom Labels. A public workforce is a group of labelers from Amazon Mechanical Turk, a crowdsourcing marketplace. The company will have to pay the public workforce for each labeling task, and it may not have full control over the quality and security of the labeled data. The company will also have to manage the SageMaker training job, the model artifact, and the batch transform job, as explained in option C4.

1: Amazon SageMaker Ground Truth

2: Amazon Rekognition Custom Labels

3: Amazon SageMaker Object Detection

4: Amazon Mechanical Turk

The elbow method is a technique that we use to determine the number of centroids (k) to use in a k-means clustering algorithm. In this method, we plot the within-cluster sum of squares (WCSS) against the number of clusters (k) and look for the point where the curve bends sharply. This point is called the elbow point and it indicates that adding more clusters does not improve the model significantly. The graph in the question shows that the elbow point is at k = 4, which means that 4 is a reasonable choice for the optimal number of clusters. References:

Elbow Method for optimal value of k in KMeans: A tutorial on how to use the elbow method with Amazon SageMaker.

K-Means Clustering: A video that explains the concept and benefits of k-means clustering.

The solution C meets the requirements because it minimizes costs for compute infrastructure, maximizes the scalability of resources for training, and facilitates the use of an ML model in low-connectivity environments. The solution C involves the following steps:

Move the training data to an Amazon S3 bucket. This will enable the company to store the large amount of data in a durable, scalable, and cost-effective way. It will also allow the company to access the data from the cloud for training and evaluation purposes1.

Train and evaluate the model by using Amazon SageMaker. This will enable the company to use a fully managed service that provides various features and tools for building, training, tuning, and deploying ML models. Amazon SageMaker can handle large-scale data processing and distributed training, and it can leverage the power of AWS compute resources such as Amazon EC2, Amazon EKS, and AWS Fargate2.

Optimize the model by using SageMaker Neo. This will enable the company to reduce the size of the model and improve its performance and efficiency. SageMaker Neo can compile the model into an executable that can run on various hardware platforms, such as CPUs, GPUs, and edge devices3.

Set up an edge device in the manufacturing facilities with AWS IoT Greengrass. This will enable the company to deploy the model on a local device that can run inference in real time, even in low-connectivity environments. AWS IoT Greengrass can extend AWS cloud capabilities to the edge, and it can securely communicate with the cloud for updates and synchronization4.

Deploy the model on the edge device. This will enable the company to automate quality control in its facilities by using the model to detect defects in new parts as they move on a conveyor belt. The model can run inference locally on the edge device without requiring internet connectivity, and it can send the results to the cloud when the connection is available4.

The other options are not suitable because:

Option A: Deploying the model on a SageMaker hosting services endpoint will not facilitate the use of the model in low-connectivity environments, as it will require internet access to perform inference. Moreover, it may incur higher costs for hosting and data transfer than deploying the model on an edge device.

Option B: Training and evaluating the model on premises will not minimize costs for compute infrastructure, as it will require the company to maintain and upgrade its own hardware and software. Moreover, it will not maximize the scalability of resources for training, as it will limit the company’s ability to leverage the cloud’s elasticity and flexibility.

Option D: Training the model on premises will not minimize costs for compute infrastructure, nor maximize the scalability of resources for training, for the same reasons as option B.

1: Amazon S3

2: Amazon SageMaker

3: SageMaker Neo

4: AWS IoT Greengrass

The factors that will adversely impact the performance of the forecast model are:

Sales data is aggregated by week. This will reduce the granularity and resolution of the data, and make it harder to capture the daily patterns and variations in sales volume. The data scientist should request daily sales data from the source database to enable building a daily model, which will be more accurate and useful for the prediction task.

Sales data is missing zero entries for item sales. This will introduce bias and incompleteness in the data, and make it difficult to account for the items that have no demand or are out of stock. The data scientist should request that item sales data from the source database include zero entries to enable building the model, which will be more robust and realistic.

The other options are not valid because:

Detecting seasonality for the majority of stores will not be an issue, as sales data is highly consistent from week to week. Requesting categorical data to relate new stores with similar stores that have more historical data may not improve the model performance significantly, and may introduce unnecessary complexity and noise.

The sales data does not need to have more variance, as it reflects the actual demand and behavior of the customers. Requesting external sales data from other industries will not improve the model’s ability to generalize, but may introduce irrelevant and misleading information.

Only 100 MB of sales data is not a problem, as it is sufficient to train a forecast model with Amazon S3 and Amazon Forecast. Requesting 10 years of sales data will not provide much benefit, as it may contain outdated and obsolete information that does not reflect the current market trends and customer preferences.

Amazon Forecast

Forecasting: Principles and Practice

Option A is correct because SageMaker Pipelines is a service that enables you to create and manage automated workflows for your machine learning projects. You can use SageMaker Pipelines to orchestrate the steps of data extraction, model training, and model deployment in a repeatable and scalable way1.

Option B is correct because SageMaker Model Monitor is a service that monitors the quality of your models in production and alerts you when there are deviations in the model quality. You can use SageMaker Model Monitor to set an accuracy threshold for your model and configure a CloudWatch alarm that triggers when the threshold is exceeded. You can then connect the alarm to the workflow in SageMaker Pipelines to automatically initiate retraining and deployment of a new version of the model2.

Option C is incorrect because it is not the most operationally efficient way to maintain the model’s accuracy. Creating a daily SageMaker Processing job that reads the predictions from Amazon S3 and checks for changes in model prediction accuracy is a manual and time-consuming process. It also requires you to write custom code to perform the data analysis and send the email notification. Moreover, it does not automatically retrain and deploy the model when the accuracy drops.

Option D is incorrect because it is not the most operationally efficient way to maintain the model’s accuracy. Rerunning the steps in the Jupyter notebook that is hosted on SageMaker Studio notebooks to retrain the model and redeploy a new version of the model is a manual and error-prone process. It also requires you to monitor the model’s performance and initiate the retraining and deployment steps yourself. Moreover, it does not leverage the benefits of SageMaker Pipelines and SageMaker Model Monitor to automate and streamline the workflow.

Option E is incorrect because it is not the most operationally efficient way to maintain the model’s accuracy. Exporting the training and deployment code from the SageMaker Studio notebooks into a Python script and packaging the script into an Amazon ECS task that an AWS Lambda function can initiate is a complex and cumbersome process. It also requires you to manage the infrastructure and resources for the Amazon ECS task and the AWS Lambda function. Moreover, it does not leverage the benefits of SageMaker Pipelines and SageMaker Model Monitor to automate and streamline the workflow.

1: SageMaker Pipelines - Amazon SageMaker

2: Monitor data and model quality - Amazon SageMaker

The solution C will improve the computational efficiency of the models because it uses Amazon SageMaker Debugger and pruning, which are techniques that can reduce the size and complexity of the convolutional neural network (CNN) models. The solution C involves the following steps:

Use Amazon SageMaker Debugger to gain visibility into the training weights, gradients, biases, and activation outputs. Amazon SageMaker Debugger is a service that can capture and analyze the tensors that are emitted during the training process of machine learning models. Amazon SageMaker Debugger can provide insights into the model performance, quality, and convergence. Amazon SageMaker Debugger can also help to identify and diagnose issues such as overfitting, underfitting, vanishing gradients, and exploding gradients1.

Compute the filter ranks based on the training information. Filter ranking is a technique that can measure the importance of each filter in a convolutional layer based on some criterion, such as the average percentage of zero activations or the L1-norm of the filter weights. Filter ranking can help to identify the filters that have little or no contribution to the model output, and thus can be removed without affecting the model accuracy2.

Apply pruning to remove the low-ranking filters. Pruning is a technique that can reduce the size and complexity of a neural network by removing the redundant or irrelevant parts of the network, such as neurons, connections, or filters. Pruning can help to improve the computational efficiency, memory usage, and inference speed of the model, as well as to prevent overfitting and improve generalization3.

Set the new weights based on the pruned set of filters. After pruning, the model will have a smaller and simpler architecture, with fewer filters in each convolutional layer. The new weights of the model can be set based on the pruned set of filters, either by initializing them randomly or by fine-tuning them from the original weights4.

Run a new training job with the pruned model. The pruned model can be trained again with the same or a different dataset, using the same or a different framework or algorithm. The new training job can use the same or a different configuration of Amazon SageMaker, such as the instance type, the hyperparameters, or the data ingestion mode. The new training job can also use Amazon SageMaker Debugger to monitor and analyze the training process and the model quality5.

The other options are not suitable because:

Option A: Using Amazon CloudWatch metrics to gain visibility into the SageMaker training weights, gradients, biases, and activation outputs will not be as effective as using Amazon SageMaker Debugger. Amazon CloudWatch is a service that can monitor and observe the operational health and performance of AWS resources and applications. Amazon CloudWatch can provide metrics, alarms, dashboards, and logs for various AWS services, including Amazon SageMaker. However, Amazon CloudWatch does not provide the same level of granularity and detail as Amazon SageMaker Debugger for the tensors that are emitted during the training process of machine learning models. Amazon CloudWatch metrics are mainly focused on the resource utilization and the training progress, not on the model performance, quality, and convergence6.

Option B: Using Amazon SageMaker Ground Truth to build and run data labeling workflows and collecting a larger labeled dataset with the labeling workflows will not improve the computational efficiency of the models. Amazon SageMaker Ground Truth is a service that can create high-quality training datasets for machine learning by using human labelers. A larger labeled dataset can help to improve the model accuracy and generalization, but it will not reduce the memory, battery, and hardware consumption of the model. Moreover, a larger labeled dataset may increase the training time and cost of the model7.

Option D: Using Amazon SageMaker Model Monitor to gain visibility into the ModelLatency metric and OverheadLatency metric of the model after the company deploys the model and increasing the model learning rate will not improve the computational efficiency of the models. Amazon SageMaker Model Monitor is a service that can monitor and analyze the quality and performance of machine learning models that are deployed on Amazon SageMaker endpoints. The ModelLatency metric and the OverheadLatency metric can measure the inference latency of the model and the endpoint, respectively. However, these metrics do not provide any information about the training weights, gradients, biases, and activation outputs of the model, which are needed for pruning. Moreover, increasing the model learning rate will not reduce the size and complexity of the model, but it may affect the model convergence and accuracy.

1: Amazon SageMaker Debugger

2: Pruning Convolutional Neural Networks for Resource Efficient Inference

3: Pruning Neural Networks: A Survey

4: Learning both Weights and Connections for Efficient Neural Networks

5: Amazon SageMaker Training Jobs

6: Amazon CloudWatch Metrics for Amazon SageMaker

7: Amazon SageMaker Ground Truth

Amazon SageMaker Model Monitor

The solution that will meet the requirements with the least amount of effort is to use a semantic segmentation algorithm to identify a visitor’s hair in video frames, and pass the identified hair to an ResNet-50 algorithm to determine hair style and hair color. This solution can leverage the existing Amazon SageMaker algorithms and frameworks to perform the tasks of hair segmentation and classification.

Semantic segmentation is a computer vision technique that assigns a class label to every pixel in an image, such that pixels with the same label share certain characteristics. Semantic segmentation can be used to identify and isolate different objects or regions in an image, such as a visitor’s hair in a video frame. Amazon SageMaker provides a built-in semantic segmentation algorithm that can train and deploy models for semantic segmentation tasks. The algorithm supports three state-of-the-art network architectures: Fully Convolutional Network (FCN), Pyramid Scene Parsing Network (PSP), and DeepLab v3. The algorithm can also use pre-trained or randomly initialized ResNet-50 or ResNet-101 as the backbone network. The algorithm can be trained using P2/P3 type Amazon EC2 instances in single machine configurations1.

ResNet-50 is a convolutional neural network that is 50 layers deep and can classify images into 1000 object categories. ResNet-50 is trained on more than a million images from the ImageNet database and can achieve high accuracy on various image recognition tasks. ResNet-50 can be used to determine hair style and hair color from the segmented hair regions in the video frames. Amazon SageMaker provides a built-in image classification algorithm that can use ResNet-50 as the network architecture. The algorithm can also perform transfer learning by fine-tuning the pre-trained ResNet-50 model with new data. The algorithm can be trained using P2/P3 type Amazon EC2 instances in single or multiple machine configurations2.

The other options are either less effective or more complex to implement. Using an object detection algorithm to identify a visitor’s hair in video frames would not segment the hair at the pixel level, but only draw bounding boxes around the hair regions. This could result in inaccurate or incomplete hair segmentation, especially if the hair is occluded or has irregular shapes. Using an XGBoost algorithm to determine hair style and hair color would require transforming the segmented hair images into numerical features, which could lose some information or introduce noise. XGBoost is also not designed for image classification tasks, and may not achieve high accuracy or performance.

1: Semantic Segmentation Algorithm - Amazon SageMaker

2: Image Classification Algorithm - Amazon SageMaker

The algorithms that are best suited to this scenario are latent Dirichlet allocation (LDA) and neural topic modeling (NTM), as they are both unsupervised learning methods that can discover abstract topics from a collection of text documents. LDA and NTM can provide a list of the top words for each topic, as well as the topic distribution for each document, which can help the auditors assess the relevance and priority of the topic12.

The other options are not suitable because:

Option B: A random forest classifier is a supervised learning method that can perform classification or regression tasks by using an ensemble of decision trees. A random forest classifier is not suitable for discovering abstract topics from text documents, as it requires labeled data and predefined classes3.

Option D: A linear support vector machine is a supervised learning method that can perform classification or regression tasks by using a linear function that separates the data into different classes. A linear support vector machine is not suitable for discovering abstract topics from text documents, as it requires labeled data and predefined classes4.

Option E: A linear regression is a supervised learning method that can perform regression tasks by using a linear function that models the relationship between a dependent variable and one or more independent variables. A linear regression is not suitable for discovering abstract topics from text documents, as it requires labeled data and a continuous output variable5.

1: Latent Dirichlet Allocation

2: Neural Topic Modeling

3: Random Forest Classifier

4: Linear Support Vector Machine

5: Linear Regression

The best solution is to configure Amazon EventBridge to run a predefined SageMaker pipeline to perform the transformations when a new data is detected in the S3 bucket. This solution requires the least development effort because it leverages the native integration between EventBridge and SageMaker Pipelines, which allows you to trigger a pipeline execution based on an event rule. EventBridge can monitor the S3 bucket for new data uploads and invoke the pipeline that contains the same transformations and feature engineering steps that were defined in SageMaker Data Wrangler. The pipeline can then ingest the transformed data into the online feature store for training and inference.

The other solutions are less optimal because they require more development effort and additional services. Using AWS Lambda or AWS Step Functions would require writing custom code to invoke the SageMaker pipeline and handle any errors or retries. Using Apache Airflow would require setting up and maintaining an Airflow server and DAGs, as well as integrating with the SageMaker API.

Amazon EventBridge and Amazon SageMaker Pipelines integration

Create a pipeline using a JSON specification

Ingest data into a feature group

 The XGBoost model is a supervised machine learning algorithm, which means it requires labeled data to learn from. The customers’ current segmentation is unknown, so there is no label to train the XGBoost model on. Moreover, the XGBoost model is designed for classification or regression tasks, not for clustering. Clustering is a type of unsupervised machine learning, which means it does not require labeled data. Clustering algorithms try to find natural groups or clusters in the data based on their similarity or distance. A common clustering algorithm is k-means, which partitions the data into K clusters, where each data point belongs to the cluster with the nearest mean. To meet the current requirements, the data scientist should train a k-means model with K = 5 on the same dataset to predict a segment for each customer. This way, the data scientist can identify five customer segments based on age, income, and location, without needing any labels. References:

What is XGBoost? - Amazon SageMaker

What is Clustering? - Amazon SageMaker

K-Means Algorithm - Amazon SageMaker

Serverless inference in SageMaker is designed for workloads with intermittent traffic and unpredictable usage patterns, which aligns with the media company's periodic high-traffic windows. Because the payload is less than 4 MB (serverless inference supports payloads up to 4 MB), serverless inference is appropriate and provisioned concurrency ensures the endpoints are warm and ready during peak times to minimize latency.

From AWS documentation:

“Amazon SageMaker Serverless Inference is ideal for applications with intermittent or unpredictable traffic. You can optionally enable provisioned concurrency to ensure that your endpoints are always ready to process requests during anticipated peak traffic hours.”

— AWS SageMaker Serverless Inference documentation

Using real-time inference with auto scaling (A) incurs a higher cost as it requires always-on instances. Batch transform (D) is for offline, large-scale inferences, not low-latency real-time inference. Asynchronous inference (C) is typically used for large payloads (>4 MB) or long processing times.

The solution that uses Amazon Comprehend and Amazon CloudWatch is the most suitable for the given scenario. Amazon Comprehend is a natural language processing (NLP) service that can analyze text and extract insights such as sentiment, entities, topics, and syntax. Amazon CloudWatch is a monitoring and observability service that can collect and track metrics, create dashboards, and set alarms based on various thresholds. By using these services, the data science team can quickly and easily implement a solution to monitor the sentiment of social media posts without requiring much infrastructure or data science resources. The solution also meets the requirements of storing the sentiment in both S3 and CloudWatch, and using CloudWatch alarms to notify analysts of trends.

Amazon Comprehend

Amazon CloudWatch

The MAPE of the predictor could be improved by making the following changes to the CreatePredictor API call:

Set PerformAutoML to true. This will allow Amazon Forecast to automatically evaluate different algorithms and choose the one that minimizes the objective function, which is the mean of the weighted losses over the forecast types. By default, these are the p10, p50, and p90 quantile losses1. This option can help find a better algorithm than DeepAR+ for the given data.

Set PerformHPO to true. This will enable hyperparameter optimization (HPO), which is the process of finding the optimal values for the algorithm-specific parameters that affect the quality of the forecasts. HPO can improve the accuracy of the predictor by tuning the hyperparameters based on the training data2.

The other options are not likely to improve the MAPE of the predictor. Setting ForecastHorizon to 4 will reduce the number of time steps that the model predicts, which may not match the business requirement of predicting next month’s inventory. Setting ForecastFrequency to W for weekly will change the granularity of the forecasts, which may not be appropriate for the monthly data. Setting FeaturizationMethodName to filling will not have any effect, since there is no missing data in the dataset.

CreatePredictor - Amazon Forecast

HPOConfig - Amazon Forecast

In Amazon SageMaker, hyperparameter tuning jobs optimize model performance by adjusting hyperparameters. Amazon SageMaker’s hyperparameter tuning supports completion criteria settings that enable efficient management of tuning resources. In this scenario, the ML specialist aims to set a completion criterion that will terminate the tuning job as soon as SageMaker detects that further improvements in the objective metric are unlikely to exceed 1%.

The CompleteOnConvergence setting is designed for such requirements. This criterion enables the tuning job to automatically stop when SageMaker determines that additional hyperparameter evaluations are unlikely to improve the objective metric beyond a certain threshold, allowing for efficient tuning completion. The convergence process relies on an internal optimization algorithm that continuously evaluates the objective metric during tuning and stops when performance stabilizes without further improvement.

This is supported by AWS documentation, which explains that CompleteOnConvergence is an efficient way to manage tuning by stopping unnecessary evaluations once the model performance stabilizes within the specified threshold.

Principal Component Analysis (PCA) is a dimensionality reduction technique that captures the variance within the feature space, helping to understand the directions in which data varies most. In SageMaker Data Wrangler, the multicollinearity measurement and PCA features allow the data scientist to analyze interdependencies between predictor variables while reducing redundancy. PCA transforms correlated features into a set of uncorrelated components, helping to simplify the dataset without significant loss of information, making it ideal for refining features based on variance.

Options A and D offer methods to understand feature relevance but are less effective for managing multicollinearity and variance representation in the data.

The solution that will accomplish the necessary transformation to train the Amazon SageMaker model with the least amount of administrative overhead is to crawl the data using AWS Glue crawlers, write an AWS Glue ETL job that merges the two tables and writes the output in CSV format to Amazon S3. This solution leverages the serverless capabilities of AWS Glue to automatically discover the schema of the data sources, and to perform the data integration and transformation without requiring any cluster management or configuration. The output in CSV format is compatible with Amazon SageMaker and can be easily loaded into a training job. References: AWS Glue, Amazon SageMaker

 The ML specialist should run the Model Monitor baseline job again on the new training set and configure Model Monitor to use the new baseline. This is because the baseline job computes the statistics and constraints for the data quality and model quality metrics, which are used to detect violations. If the training set changes, the baseline job should be updated accordingly to reflect the new distribution of the data and the model performance. Otherwise, the old baseline may not be representative of the current production traffic and may cause false alarms or miss violations. References:

Monitor data and model quality - Amazon SageMaker

Detecting and analyzing incorrect model predictions with Amazon SageMaker Model Monitor and Debugger | AWS Machine Learning Blog