Free Amazon MLA-C01 Übungsprüfungen - Seite 3 von 6

Question #21

What is the first step in building a machine learning application?

A . Preparing the data
B . Training the model
C . Formulating the problem
D . Testing the model

Question #22

You are a data scientist at a marketing agency tasked with creating a sentiment analysis model to analyze customer reviews for a new product. The company wants to quickly deploy a solution with minimal training time and development effort. You decide to leverage a pre-trained natural language processing (NLP) model and fine-tune it using a custom dataset of labeled customer reviews. Your team has access to both Amazon Bedrock and SageMaker JumpStart.

Which approach is the MOST APPROPRIATE for fine-tuning the pre-trained model with your custom dataset?

A . Use SageMaker JumpStart to create a custom container for your pre-trained model and manually implement fine-tuning with TensorFlow
B . Use SageMaker JumpStart to deploy a pre-trained NLP model and use the built-in fine-tuning functionality with your custom dataset to create a customized sentiment analysis model
C . Use Amazon Bedrock to train a model from scratch using your custom dataset, as Bedrock is optimized for training large models efficiently
D . Use Amazon Bedrock to select a foundation model from a third-party provider, then fine-tune the
model directly in the Bedrock interface using your custom dataset

Lösung einblenden Lösung ausblenden

Richtige Antwort: B
B

Explanation:

Correct option:

Use SageMaker JumpStart to deploy a pre-trained NLP model and use the built-in fine-tuning functionality with your custom dataset to create a customized sentiment analysis model

Amazon Bedrock is the easiest way to build and scale generative AI applications with foundation models. Amazon Bedrock is a fully managed service that offers a choice of high-performing foundation models (FMs) from leading AI companies like AI21 Labs, Anthropic, Cohere, Meta, Mistral AI, Stability AI, and

Amazon through a single API, along with a broad set of capabilities you need to build generative AI applications with security, privacy, and responsible AI.

Amazon SageMaker JumpStart is a machine learning (ML) hub that can help you accelerate your ML journey. With SageMaker JumpStart, you can evaluate, compare, and select FMs quickly based on pre-defined quality and responsibility metrics to perform tasks like article summarization and image generation. SageMaker JumpStart provides managed infrastructure and tools to accelerate scalable, reliable, and secure model building, training, and deployment of ML models.

Fine-tuning trains a pretrained model on a new dataset without training from scratch. This process, also known as transfer learning, can produce accurate models with smaller datasets and less training time.

SageMaker JumpStart is specifically designed for scenarios like this, where you can quickly deploy a pre-trained model and fine-tune it using your custom dataset. This approach allows you to leverage existing NLP models, reducing both development time and computational resources needed for training from scratch.

via – https://docs.aws.amazon.com/sagemaker/latest/dg/jumpstart-fine-tune.html

Incorrect options:

Use Amazon Bedrock to select a foundation model from a third-party provider, then fine-tune the model directly in the Bedrock interface using your custom dataset – Amazon Bedrock provides access to foundation models from third-party providers, allowing for easy deployment and integration into

applications. However, as of now, Bedrock does not directly support fine-tuning these models within its interface. Fine-tuning is better suited for SageMaker JumpStart in this scenario.

Use Amazon Bedrock to train a model from scratch using your custom dataset, as Bedrock is optimized for training large models efficiently – Amazon Bedrock is not intended for training models from scratch, especially not for scenarios where fine-tuning a pre-trained model would be more efficient. Bedrock is optimized for deploying and scaling foundation models, not for raw model training.

Use SageMaker JumpStart to create a custom container for your pre-trained model and manually implement fine-tuning with TensorFlow – While it’s possible to create a custom container and manually fine-tune a model, SageMaker JumpStart already offers an integrated solution for fine-tuning pre-trained models without the need for custom containers or manual implementation. This makes it a more efficient and straightforward option for the task at hand.

Reference: https://docs.aws.amazon.com/sagemaker/latest/dg/jumpstart-fine-tune.html

Question #23

How do financial institutions use machine learning with mobile check deposits?

A . To automate logistics
B . To improve ride-sharing apps
C . To reduce wait times
D . To recognize content

Lösung einblenden Lösung ausblenden

Question #24

You are a machine learning engineer working for an e-commerce company. You have developed a recommendation model that predicts products customers are likely to buy based on their browsing

history and past purchases. The model initially performs well, but after deploying it in production, you notice two issues: the model’s performance degrades over time as new data is added (catastrophic forgetting) and the model shows signs of overfitting during retraining on updated datasets.

Given these challenges, which of the following strategies is the MOST LIKELY to help prevent overfitting and catastrophic forgetting while maintaining model accuracy?

A . Reduce the model complexity by decreasing the number of features, apply data augmentation to handle underfitting, and leverage L1 regularization to address catastrophic forgetting
B . Apply L2 regularization to reduce overfitting, use dropout to prevent underfitting, and retrain the model on the entire dataset periodically to avoid catastrophic forgetting
C . Use early stopping during training to prevent overfitting, incorporate new data incrementally through transfer learning to mitigate catastrophic forgetting, and apply L1 regularization to ensure feature selection
D . Regularly update the training dataset with new data, apply L2 regularization to manage overfitting,
and use an ensemble of models to prevent catastrophic forgetting

Lösung einblenden Lösung ausblenden

Richtige Antwort: C
C

Explanation:

Correct option:

Use early stopping during training to prevent overfitting, incorporate new data incrementally through transfer learning to mitigate catastrophic forgetting, and apply L1 regularization to ensure feature selection

Early stopping is a proven method to prevent overfitting by halting training when the model’s performance on the validation set stops improving. Incorporating new data incrementally through transfer learning helps to mitigate catastrophic forgetting by allowing the model to learn new information while retaining its prior knowledge.

via – https://docs.aws.amazon.com/sagemaker/latest/dg/automatic-model-tuning-early-stopping.html

Regularization helps prevent linear models from overfitting training data examples by penalizing extreme weight values. L1 regularization reduces the number of features used in the model by pushing the weight of features that would otherwise have very small weights to zero. L1 regularization produces sparse models and reduces the amount of noise in the model. L2 regularization results in smaller overall weight values, which stabilizes the weights when there is high correlation between the features.

L1 regularization is beneficial for feature selection, which can improve model generalization and prevent both overfitting and underfitting.

via – https://aws.amazon.com/blogs/machine-learning/automatically-retrain-neural-networks-with-renate/

Incorrect options:

Apply L2 regularization to reduce overfitting, use dropout to prevent underfitting, and retrain the model on the entire dataset periodically to avoid catastrophic forgetting – While L2 regularization is effective for reducing overfitting, using dropout typically addresses overfitting, not underfitting. Retraining the model on the entire dataset periodically may help, but it could still lead to catastrophic forgetting as the model might "forget" previous knowledge when learning from new data.

Reduce the model complexity by decreasing the number of features, apply data augmentation to handle

underfitting, and leverage L1 regularization to address catastrophic forgetting – Reducing model complexity can help with overfitting, but it might lead to underfitting if too many features are removed. Data augmentation is more relevant for addressing overfitting, particularly in image or text data, rather than underfitting. L1 regularization is effective for reducing overfitting, and not for addressing catastrophic forgetting.

Regularly update the training dataset with new data, apply L2 regularization to manage overfitting, and use an ensemble of models to prevent catastrophic forgetting – Regularly updating the dataset can help with model performance over time, but it might not address catastrophic forgetting unless combined with techniques like transfer learning. L2 regularization and ensembling are useful strategies, but ensembling does not specifically prevent catastrophic forgetting.

References:

https://aws.amazon.com/blogs/machine-learning/automatically-retrain-neural-networks-with-renate/

https://docs.aws.amazon.com/machine-learning/latest/dg/training-parameters.html

https://docs.aws.amazon.com/sagemaker/latest/dg/automatic-model-tuning-early-stopping.html

Question #24

You are a machine learning engineer working for an e-commerce company. You have developed a recommendation model that predicts products customers are likely to buy based on their browsing

history and past purchases. The model initially performs well, but after deploying it in production, you notice two issues: the model’s performance degrades over time as new data is added (catastrophic forgetting) and the model shows signs of overfitting during retraining on updated datasets.

Given these challenges, which of the following strategies is the MOST LIKELY to help prevent overfitting and catastrophic forgetting while maintaining model accuracy?

A . Reduce the model complexity by decreasing the number of features, apply data augmentation to handle underfitting, and leverage L1 regularization to address catastrophic forgetting
B . Apply L2 regularization to reduce overfitting, use dropout to prevent underfitting, and retrain the model on the entire dataset periodically to avoid catastrophic forgetting
C . Use early stopping during training to prevent overfitting, incorporate new data incrementally through transfer learning to mitigate catastrophic forgetting, and apply L1 regularization to ensure feature selection
D . Regularly update the training dataset with new data, apply L2 regularization to manage overfitting,
and use an ensemble of models to prevent catastrophic forgetting

Lösung einblenden Lösung ausblenden

Richtige Antwort: C
C

Explanation:

Correct option:

Use early stopping during training to prevent overfitting, incorporate new data incrementally through transfer learning to mitigate catastrophic forgetting, and apply L1 regularization to ensure feature selection

Early stopping is a proven method to prevent overfitting by halting training when the model’s performance on the validation set stops improving. Incorporating new data incrementally through transfer learning helps to mitigate catastrophic forgetting by allowing the model to learn new information while retaining its prior knowledge.

via – https://docs.aws.amazon.com/sagemaker/latest/dg/automatic-model-tuning-early-stopping.html

Regularization helps prevent linear models from overfitting training data examples by penalizing extreme weight values. L1 regularization reduces the number of features used in the model by pushing the weight of features that would otherwise have very small weights to zero. L1 regularization produces sparse models and reduces the amount of noise in the model. L2 regularization results in smaller overall weight values, which stabilizes the weights when there is high correlation between the features.

L1 regularization is beneficial for feature selection, which can improve model generalization and prevent both overfitting and underfitting.

via – https://aws.amazon.com/blogs/machine-learning/automatically-retrain-neural-networks-with-renate/

Incorrect options:

Apply L2 regularization to reduce overfitting, use dropout to prevent underfitting, and retrain the model on the entire dataset periodically to avoid catastrophic forgetting – While L2 regularization is effective for reducing overfitting, using dropout typically addresses overfitting, not underfitting. Retraining the model on the entire dataset periodically may help, but it could still lead to catastrophic forgetting as the model might "forget" previous knowledge when learning from new data.

Reduce the model complexity by decreasing the number of features, apply data augmentation to handle

underfitting, and leverage L1 regularization to address catastrophic forgetting – Reducing model complexity can help with overfitting, but it might lead to underfitting if too many features are removed. Data augmentation is more relevant for addressing overfitting, particularly in image or text data, rather than underfitting. L1 regularization is effective for reducing overfitting, and not for addressing catastrophic forgetting.

Regularly update the training dataset with new data, apply L2 regularization to manage overfitting, and use an ensemble of models to prevent catastrophic forgetting – Regularly updating the dataset can help with model performance over time, but it might not address catastrophic forgetting unless combined with techniques like transfer learning. L2 regularization and ensembling are useful strategies, but ensembling does not specifically prevent catastrophic forgetting.

References:

https://aws.amazon.com/blogs/machine-learning/automatically-retrain-neural-networks-with-renate/

https://docs.aws.amazon.com/machine-learning/latest/dg/training-parameters.html

https://docs.aws.amazon.com/sagemaker/latest/dg/automatic-model-tuning-early-stopping.html

Question #26

You are a Data Scientist working for a retail company and you have been tasked with developing a demand forecasting model using Amazon SageMaker. The company requires the model to be highly accurate to optimize inventory levels, but you also need to consider constraints on training time and cost due to budget limitations. You have access to multiple SageMaker instance types and options like spot instances to reduce costs, but you must balance these factors against the need for a performant model. Your goal is to choose a configuration that provides an acceptable tradeoff between model performance, training time, and cost.

Which of the following strategies should you consider when balancing model performance, training time, and cost in this scenario using Amazon SageMaker? (Select two)

A . Implement distributed training across multiple smaller instances to balance training time and cost while maintaining model performance
B . Deploy multiple models with different instance types simultaneously, choosing the one that completes training first, regardless of performance or cost
C . Optimize hyperparameters using Amazon SageMaker’s automatic model tuning (hyperparameter optimization) to improve performance, while using spot instances to reduce cost
D . Use a smaller instance type to save on costs, and accept longer training times, as model accuracy is not the highest priority
E . Use the largest available instance type to minimize training time, regardless of cost, ensuring that the model is trained as quickly as possible

Lösung einblenden Lösung ausblenden

Richtige Antwort: A, C
A, C

Explanation:

Correct options:

Optimize hyperparameters using Amazon SageMaker’s automatic model tuning (hyperparameter optimization) to improve performance, while using spot instances to reduce cost

Amazon SageMaker’s automatic model tuning allows you to optimize hyperparameters, which can significantly improve model performance without manually tuning each parameter. By using spot instances, you can reduce the cost of training while still focusing on achieving high performance. This approach helps balance the tradeoff between cost and performance effectively.

via – https://docs.aws.amazon.com/sagemaker/latest/dg/automatic-model-tuning.html

Implement distributed training across multiple smaller instances to balance training time and cost while maintaining model performance

Distributed training across multiple smaller instances can help reduce overall training time while managing costs. This strategy allows you to leverage parallel processing without incurring the high costs associated with using a single, large instance. It also ensures that the model training process is efficient without compromising on performance.

via – https://docs.aws.amazon.com/sagemaker/latest/dg/distributed-training.html

Incorrect options:

Use the largest available instance type to minimize training time, regardless of cost, ensuring that the model is trained as quickly as possible – Using the largest available instance type might minimize training time but can result in prohibitively high costs, which is not suitable given the budget constraints. This strategy does not balance cost-effectiveness with the need for performance.

Use a smaller instance type to save on costs, and accept longer training times, as model accuracy is not the highest priority – While using a smaller instance type may save on costs, it can significantly increase training time. Given that model accuracy is important for optimizing inventory levels, this tradeoff might not be acceptable.

Deploy multiple models with different instance types simultaneously, choosing the one that completes training first, regardless of performance or cost – Deploying multiple models simultaneously on different instance types and selecting the one that finishes first is inefficient and can lead to unnecessary costs. Additionally, this approach does not consider the need for balanced performance, cost, and training time.

References:

https://docs.aws.amazon.com/sagemaker/latest/dg/automatic-model-tuning.html

https://docs.aws.amazon.com/sagemaker/latest/dg/distributed-training.html

Question #27

You are a DevOps engineer at a tech company that is building a scalable microservices-based application. The application is composed of several containerized services, each responsible for different parts of the application, such as user authentication, data processing, and recommendation systems. The company wants to standardize and automate the deployment and management of its infrastructure using Infrastructure as Code (IaC). You need to choose between AWS CloudFormation and AWS Cloud Development Kit (CDK) for defining the infrastructure. Additionally, you must decide on the appropriate AWS container service to manage and deploy these microservices efficiently.

Given the requirements, which combination of IaC option and container service is MOST SUITABLE for this scenario, and why?

A . Use AWS CloudFormation with YAML templates for infrastructure automation and deploy the containerized microservices using Amazon Lightsail Containers to simplify management and reduce costs
B . Use AWS CloudFormation to define and deploy the infrastructure as code, and Amazon ECR (Elastic Container Registry) with Fargate for running the containerized microservices without needing to manage the underlying servers
C . Use AWS CDK for infrastructure as code, allowing you to define the infrastructure in a high-level programming language, and deploy the containerized microservices using Amazon EKS (Elastic Kubernetes Service) for advanced orchestration and scalability
D . Use AWS CDK with Amazon ECS on EC2 instances to combine the flexibility of programming languages with direct control over the underlying server infrastructure for the microservices

Lösung einblenden Lösung ausblenden

Richtige Antwort: C
C

Explanation:

Correct option:

Use AWS CDK for infrastructure as code, allowing you to define the infrastructure in a high-level programming language, and deploy the containerized microservices using Amazon EKS (Elastic Kubernetes Service) for advanced orchestration and scalability

AWS CDK offers the flexibility of using high-level programming languages (e.g., Python, JavaScript) to define infrastructure, making it easier to manage complex infrastructure setups programmatically. via – https://docs.aws.amazon.com/cdk/v2/guide/home.html

Amazon EKS is designed for running containerized microservices with Kubernetes, providing advanced orchestration, scalability, and integration with CI/CD pipelines. This combination is ideal for a microservices-based application with complex deployment and scaling needs.

via – https://docs.aws.amazon.com/eks/latest/userguide/what-is-eks.html

Incorrect options:

Use AWS CloudFormation to define and deploy the infrastructure as code, and Amazon ECR (Elastic Container Registry) with Fargate for running the containerized microservices without needing to manage the underlying servers – AWS CloudFormation is powerful for defining infrastructure using JSON or YAML. However, Amazon ECR is an AWS managed container image registry service and it cannot be used to manage and run containers.

Use AWS CloudFormation with YAML templates for infrastructure automation and deploy the containerized microservices using Amazon Lightsail Containers to simplify management and reduce costs – AWS CloudFormation with YAML templates is suitable for traditional IaC, but Amazon Lightsail Containers is better for simple, low-cost container deployments. It may lack the scalability and orchestration features required for a complex microservices architecture.

Use AWS CDK with Amazon ECS on EC2 instances to combine the flexibility of programming languages with direct control over the underlying server infrastructure for the microservices – AWS CDK combined with Amazon ECS on EC2 gives more control over the underlying infrastructure but adds complexity in managing the servers. For a microservices-based application, this might introduce unnecessary overhead compared to using managed services like Fargate or EKS.

References:

https://docs.aws.amazon.com/cdk/v2/guide/home.html https://docs.aws.amazon.com/eks/latest/userguide/what-is-eks.html

Question #27

You are a DevOps engineer at a tech company that is building a scalable microservices-based application. The application is composed of several containerized services, each responsible for different parts of the application, such as user authentication, data processing, and recommendation systems. The company wants to standardize and automate the deployment and management of its infrastructure using Infrastructure as Code (IaC). You need to choose between AWS CloudFormation and AWS Cloud Development Kit (CDK) for defining the infrastructure. Additionally, you must decide on the appropriate AWS container service to manage and deploy these microservices efficiently.

Given the requirements, which combination of IaC option and container service is MOST SUITABLE for this scenario, and why?

A . Use AWS CloudFormation with YAML templates for infrastructure automation and deploy the containerized microservices using Amazon Lightsail Containers to simplify management and reduce costs
B . Use AWS CloudFormation to define and deploy the infrastructure as code, and Amazon ECR (Elastic Container Registry) with Fargate for running the containerized microservices without needing to manage the underlying servers
C . Use AWS CDK for infrastructure as code, allowing you to define the infrastructure in a high-level programming language, and deploy the containerized microservices using Amazon EKS (Elastic Kubernetes Service) for advanced orchestration and scalability
D . Use AWS CDK with Amazon ECS on EC2 instances to combine the flexibility of programming languages with direct control over the underlying server infrastructure for the microservices

Lösung einblenden Lösung ausblenden

Richtige Antwort: C
C

Explanation:

Correct option:

Use AWS CDK for infrastructure as code, allowing you to define the infrastructure in a high-level programming language, and deploy the containerized microservices using Amazon EKS (Elastic Kubernetes Service) for advanced orchestration and scalability

AWS CDK offers the flexibility of using high-level programming languages (e.g., Python, JavaScript) to define infrastructure, making it easier to manage complex infrastructure setups programmatically. via – https://docs.aws.amazon.com/cdk/v2/guide/home.html

Amazon EKS is designed for running containerized microservices with Kubernetes, providing advanced orchestration, scalability, and integration with CI/CD pipelines. This combination is ideal for a microservices-based application with complex deployment and scaling needs.

via – https://docs.aws.amazon.com/eks/latest/userguide/what-is-eks.html

Incorrect options:

Use AWS CloudFormation to define and deploy the infrastructure as code, and Amazon ECR (Elastic Container Registry) with Fargate for running the containerized microservices without needing to manage the underlying servers – AWS CloudFormation is powerful for defining infrastructure using JSON or YAML. However, Amazon ECR is an AWS managed container image registry service and it cannot be used to manage and run containers.

Use AWS CloudFormation with YAML templates for infrastructure automation and deploy the containerized microservices using Amazon Lightsail Containers to simplify management and reduce costs – AWS CloudFormation with YAML templates is suitable for traditional IaC, but Amazon Lightsail Containers is better for simple, low-cost container deployments. It may lack the scalability and orchestration features required for a complex microservices architecture.

Use AWS CDK with Amazon ECS on EC2 instances to combine the flexibility of programming languages with direct control over the underlying server infrastructure for the microservices – AWS CDK combined with Amazon ECS on EC2 gives more control over the underlying infrastructure but adds complexity in managing the servers. For a microservices-based application, this might introduce unnecessary overhead compared to using managed services like Fargate or EKS.

References:

https://docs.aws.amazon.com/cdk/v2/guide/home.html https://docs.aws.amazon.com/eks/latest/userguide/what-is-eks.html

Question #29

You are a machine learning engineer responsible for deploying a customer churn prediction model using Amazon SageMaker into production at a telecommunications company. The model is critical for proactive customer retention efforts, so maintaining high availability and reliability is essential. Given the model’s importance, you must implement best practices for deployment, including versioning and rollback strategies, to ensure that any issues with the new model version can be quickly addressed without impacting the business.

Which of the following approaches BEST exemplifies deployment best practices in this scenario?

A . Version the model by tagging the new version, deploy it to production, and use weighted traffic splitting to send a small percentage of traffic to the new model. If no issues are detected, gradually increase traffic to the new version
B . Use a blue/green deployment strategy to deploy the new model version in parallel with the existing version, allowing you to switch traffic to the new model gradually and roll back if necessary
C . Deploy the new model version to a staging environment, test it thoroughly, and then manually replace the existing production model. If any issues occur, update the model in staging and redeploy
D . Deploy the new model version directly to production, replacing the existing model, and monitor its
performance closely. If issues arise, retrain the model immediately and redeploy

Lösung einblenden Lösung ausblenden

Richtige Antwort: B
B

Explanation:

Correct option:

Use a blue/green deployment strategy to deploy the new model version in parallel with the existing version, allowing you to switch traffic to the new model gradually and roll back if necessary

A blue/green deployment strategy is a best practice in model deployment. It allows you to deploy the new model version in parallel with the existing one, gradually shifting traffic to the new version while monitoring its performance. If issues are detected, you can quickly roll back to the previous version without disrupting service.

In a blue/green deployment, SageMaker provisions a new fleet with the updates (the green fleet). Then, SageMaker shifts traffic from the old fleet (the blue fleet) to the green fleet. Once the green fleet operates smoothly for a set evaluation period (called the baking period), SageMaker terminates the blue fleet. You can specify Amazon CloudWatch alarms that SageMaker uses to monitor the green fleet. If an issue with the updated code trips any of the alarms, SageMaker initiates an auto-rollback to the blue fleet in order to maintain availability thereby minimizing risk.

via – https://docs.aws.amazon.com/sagemaker/latest/dg/deployment-guardrails-blue-green.html

Incorrect options:

Deploy the new model version directly to production, replacing the existing model, and monitor its performance closely. If issues arise, retrain the model immediately and redeploy – Deploying the new model directly to production without a rollback strategy is risky. If issues arise, the model would need to be retrained and redeployed, which could lead to significant downtime and business impact.

Deploy the new model version to a staging environment, test it thoroughly, and then manually replace the existing production model. If any issues occur, update the model in staging and redeploy – Deploying to a staging environment is a good practice for testing, but manually replacing the production model increases the risk of downtime and errors. A more automated and controlled strategy like blue/green deployment is preferable.

Version the model by tagging the new version, deploy it to production, and use weighted traffic splitting to send a small percentage of traffic to the new model. If no issues are detected, gradually increase traffic to the new version – Versioning and using weighted traffic splitting is also a good practice, but it’s more complex and might not provide the same immediate rollback capability as blue/green deployment. However, this approach is viable in certain scenarios where gradual rollout is necessary.

References:

https://docs.aws.amazon.com/sagemaker/latest/dg/deployment-guardrails-blue-green.html

https://docs.aws.amazon.com/sagemaker/latest/dg/deployment-guardrails-rolling.html

Question #30

You are working on a machine learning project for a financial services company, developing a model to predict credit risk. After deploying the initial version of the model using Amazon SageMaker, you find that its performance, measured by the AUC (Area Under the Curve), is not meeting the company’s accuracy

requirements. Your team has gathered more data and believes that the model can be further optimized. You are considering various methods to improve the model’s performance, including feature engineering, hyperparameter tuning, and trying different algorithms. However, given the limited time and computational resources, you need to prioritize the most impactful strategies.

Which of the following approaches are the MOST LIKELY to lead to a significant improvement in model performance? (Select two)

A . Increase the size of the training dataset by incorporating synthetic data and then retrain the existing model
B . Perform hyperparameter tuning using Bayesian optimization and increase the number of trials to explore a broader search space
C . Switch to a more complex algorithm, such as deep learning, and use transfer learning to leverage pre-trained models
D . Use Amazon SageMaker Debugger to debug and improve model performance by addressing underlying problems such as overfitting, saturated activation functions, and vanishing gradients
E . Focus on feature engineering by creating

Lösung einblenden Lösung ausblenden

Richtige Antwort: D, E
D, E

Explanation:

Correct options:

Focus on feature engineering by creating domain-specific features and use SageMaker Clarify to evaluate feature importance

Feature engineering is one of the most effective ways to boost model performance, particularly in model with better signals for prediction. SageMaker Clarify can be used to evaluate feature importance, helping you identify the most impactful features and further refine the model.

via – https://aws.amazon.com/sagemaker/clarify/

Use Amazon SageMaker Debugger to debug and improve model performance by addressing underlying problems such as overfitting, saturated activation functions, and vanishing gradients

A machine learning (ML) training job can have problems such as overfitting, saturated activation functions, and vanishing gradients, which can compromise model performance.

SageMaker Debugger provides tools to debug training jobs and resolve such problems to improve the performance of your model. Debugger also offers tools to send alerts when training anomalies are found, take actions against the problems, and identify the root cause of them by visualizing collected metrics and tensors.

SageMaker Debugger:

via – https://docs.aws.amazon.com/sagemaker/latest/dg/train-debugger.html

Incorrect options:

Increase the size of the training dataset by incorporating synthetic data and then retrain the existing model – Increasing the size of the dataset with synthetic data can improve model performance, but it also introduces the risk of adding noise or bias if the synthetic data is not carefully generated. This approach may not guarantee a significant performance boost unless the original dataset was severely lacking in size.

Switch to a more complex algorithm, such as deep learning, and use transfer learning to leverage pre-trained models – Switching to a more complex algorithm or using transfer learning could improve performance, but it also increases the risk of overfitting, especially if the new algorithm is not well suited to the data. Additionally, deep learning models require more data and tuning, which may not be feasible given the time and resource constraints.

Perform hyperparameter tuning using Bayesian optimization and increase the number of trials to explore a broader search space – Hyperparameter tuning, especially using Bayesian optimization, can help optimize the model’s performance, but the gains might be marginal if the underlying features are not informative. It’s a valuable approach, but may not be the most impactful first step.

References:

https://docs.aws.amazon.com/sagemaker/latest/dg/train-debugger.html

https://aws.amazon.com/sagemaker/clarify/