AWS Certified Machine Learning - Specialty Practice Test

Answer: B,C,E
Explanation: The correct combination of actions to enable the data scientist’s IAM user to
invoke the SageMaker endpoint is B, C, and E, because they ensure that the IAM user hasthe necessary permissions, access, and syntax to query the ML model from Athena. Theseactions have the following benefits:B: Including a policy statement for the IAM user that allows thesagemaker:InvokeEndpoint action grants the IAM user the permission to call theSageMaker Runtime InvokeEndpoint API, which is used to get inferences from themodel hosted at the endpoint1.C: Including an inline policy for the IAM user that allows SageMaker to read S3objects enables the IAM user to access the data stored in S3, which is the sourceof the Athena queries2.E: Including the SQL statement “USING EXTERNAL FUNCTIONml_function_name” in the Athena SQL query allows the IAM user to invoke the MLmodel as an external function from Athena, which is a feature that enablesquerying ML models from SQL statements3.The other options are not correct or necessary, because they have the followingdrawbacks:A: Attaching the AmazonAthenaFullAccess AWS managed policy to the useridentity is not sufficient, because it does not grant the IAM user the permission toinvoke the SageMaker endpoint, which is required to query the ML model4.D: Including a policy statement for the IAM user that allows the IAM user toperform the sagemaker:GetRecord action is not relevant, because this action isused to retrieve a single record from a feature group, which is not the case in thisscenario5.F: Performing a user remapping in SageMaker to map the IAM user to anotherIAM user that is on the hosted endpoint is not applicable, because this feature isonly available for multi-model endpoints, which are not used in this scenario.References:1: InvokeEndpoint - Amazon SageMaker2: Querying Data in Amazon S3 from Amazon Athena - Amazon Athena3: Querying machine learning models from Amazon Athena using AmazonSageMaker | AWS Machine Learning Blog 4: AmazonAthenaFullAccess - AWS Identity and Access Management5: GetRecord - Amazon SageMaker Feature Store Runtime: [Invoke a Multi-Model Endpoint - Amazon SageMaker]

Answer: B
Explanation: 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 prebuiltcontainers for various frameworks such as TensorFlow. Script mode supports reading datafrom 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 wouldmeet the business requirements with the least development overhead.Option A is incorrect because using a local path of the data would not be scalable orreliable, 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 asdurability, security, and performance. Option C is incorrect because rewriting the train.pyscript to convert TFRecords to protobuf would require additional development effort andcomplexity, 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 AmazonSageMaker would also require additional development effort and complexity, as well asinvolve using additional services such as AWS Glue or AWS Lambda, which wouldincrease the cost and maintenance of the solution.References:Bring your own model with Amazon SageMaker script modeGitHub - aws-samples/amazon-sagemaker-script-modeDeep Dive on TensorFlow training with Amazon SageMaker and Amazon S3amazon-sagemaker-script-mode/generate_cifar10_tfrecords.py at master

Answer: A
Explanation: 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 theminority 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 affectthe 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 significantlyunderrepresented in the dataset compared to the other class (usually the negative or nontargetclass). 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 themodel to be biased towards the majority class, and fail to capture the characteristics andpatterns of the minority class. As a result, the model may have a high overall accuracy, buta low recall or true positive rate for the minority class, which means it misses manyfraudulent transactions.SMOTE is a technique that can help mitigate the class imbalance problem by generatingsynthetic samples for the minority class. SMOTE works by finding the k-nearest neighborsof each minority class instance, and randomly creating new instances along the linesegments connecting them. This way, SMOTE can increase the number and diversity ofthe minority class instances, without duplicating or losing any information. By applyingSMOTE on the minority class in the training dataset, the data scientist can balance theclasses and improve the model’s performance on the positive class1.The other options are either ineffective or counterproductive. Applying SMOTE on themajority class would not balance the classes, but increase the imbalance and the size ofthe dataset. Undersampling the minority class would reduce the number of instancesavailable 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 thedataset, and introduce redundancy and overfitting.References:1: SMOTE for Imbalanced Classification with Python - Machine Learning Mastery

Answer: A,C
Explanation: The algorithms that are best suited to this scenario are latent Dirichletallocation (LDA) and neural topic modeling (NTM), as they are both unsupervised learningmethods that can discover abstract topics from a collection of text documents. LDA andNTM can provide a list of the top words for each topic, as well as the topic distribution foreach document, which can help the auditors assess the relevance and priority of thetopic12.The other options are not suitable because:Option B: A random forest classifier is a supervised learning method that canperform classification or regression tasks by using an ensemble of decisiontrees. A random forest classifier is not suitable for discovering abstract topics fromtext documents, as it requires labeled data and predefined classes3.Option D: A linear support vector machine is a supervised learning method thatcan perform classification or regression tasks by using a linear function thatseparates the data into different classes. A linear support vector machine is notsuitable for discovering abstract topics from text documents, as it requires labeleddata and predefined classes4.Option E: A linear regression is a supervised learning method that can performregression tasks by using a linear function that models the relationship between adependent variable and one or more independent variables. A linear regression isnot suitable for discovering abstract topics from text documents, as it requireslabeled data and a continuous output variable5.References:1: Latent Dirichlet Allocation2: Neural Topic Modeling3: Random Forest Classifier4: Linear Support Vector Machine5: Linear Regression

Answer: C
Explanation: The solution C will meet the requirements with the least development effortbecause it uses Amazon Rekognition and AWS CloudTrail, which are fully managedservices that can provide the desired functionality. The solution C involves the followingsteps:Use Amazon Rekognition to identify celebrities in the pictures. AmazonRekognition is a service that can analyze images and videos and extract insightssuch as faces, objects, scenes, emotions, and more. Amazon Rekognition alsoprovides a feature called Celebrity Recognition, which can recognize thousands ofcelebrities across a number of categories, such as politics, sports, entertainment,and media. Amazon Rekognition can return the name, face, and confidence scoreof the recognized celebrities, as well as additional information such as URLs andbiographies1.Use AWS CloudTrail to capture IP address and timestamp details. AWS CloudTrailis a service that can record the API calls and events made by or on behalf of AWSaccounts. AWS CloudTrail can provide information such as the source IP address,the user identity, the request parameters, and the response elements of the APIcalls. AWS CloudTrail can also deliver the event records to an Amazon S3 bucketor an Amazon CloudWatch Logs group for further analysis and auditing2.The other options are not suitable because:Option A: Using AWS Panorama to identify celebrities in the pictures and usingAWS CloudTrail to capture IP address and timestamp details will not meet therequirements effectively. AWS Panorama is a service that can extend computervision to the edge, where it can run inference on video streams from cameras andother devices. AWS Panorama is not designed for identifying celebrities inpictures, and it may not provide accurate or relevant results. Moreover, AWSPanorama requires the use of an AWS Panorama Appliance or a compatibledevice, which may incur additional costs and complexity3.Option B: Using AWS Panorama to identify celebrities in the pictures and makingcalls to the AWS Panorama Device SDK to capture IP address and timestampdetails will not meet the requirements effectively, for the same reasons as optionA. Additionally, making calls to the AWS Panorama Device SDK will require moredevelopment effort than using AWS CloudTrail, as it will involve writing customcode and handling errors and exceptions4.Option D: Using Amazon Rekognition to identify celebrities in the pictures andusing the text detection feature to capture IP address and timestamp details willnot meet the requirements effectively. The text detection feature of AmazonRekognition is used to detect and recognize text in images and videos, such asstreet names, captions, product names, and license plates. It is not suitable forcapturing IP address and timestamp details, as these are not part of the picturesthat users upload. Moreover, the text detection feature may not be accurate orreliable, as it depends on the quality and clarity of the text in the images andvideos5.References:1: Amazon Rekognition Celebrity Recognition 2: AWS CloudTrail Overview3: AWS Panorama Overview4: AWS Panorama Device SDK5: Amazon Rekognition Text Detection