In this course, you will learn how to refine machine learning (ML) models. You begin by reviewing methods for bias mitigation and model performance, and you learn how to prevent model overfitting and underfitting. Then, you will discover how to combine methods to improve model performance and how to use hyperparameter tuning to produce optimized model results.
You will also examine variations of model size and model versioning, and you will explore how Amazon SageMaker services can assist with the model refinement process.
- Course level:Advanced
- Duration: 2 hours
Activities
- Online materials
- Exercises
- Knowledge check questions
Course objectives
- Define and interpret model evaluation metrics such as model bias and model variance.
- Describe methods for detecting model overfitting and underfitting.
- Use regularization techniques and feature selection to prevent model overfitting and underfitting.
- Combine multiple training models through ensemble methods such as boosting, bagging, and stacking to improve model performance.
- Explain how hyperparameters affect model performance.
- Define key hyperparameter tuning techniques.
- Perform automated hyperparameter optimization to improve model performance.
- Identify key factors that influence model size.
- Reduce model size by using iterative model pruning.
- Use custom datasets to fine-tune pre-trained models using Amazon SageMaker JumpStart and Amazon Bedrock.
- Use regularization techniques and feature selection to prevent catastrophic forgetting.
- Manage model versions for repeatability and audit using the Amazon SageMaker Model Registry.
Intended audience
- Cloud architects
- Machine learning engineers
Recommended Skills
- Completed at least 1 year of experience using SageMaker and other AWS services for ML engineering
- Completed at least 1 year of experience in a related role such as backend software developer, DevOps developer, data engineer, or data scientist
- A fundamental understanding of programming languages such as Python
- Completed preceding courses in the AWS ML Engineer Associate Learning Plan
Course outline
- Section 1: Introduction
- Lesson 1: How to Use This Course
- Lesson 2: Course Overview
- Lesson 3: Evaluating Model Performance
- Section 2: Model Fit
- Lesson 4: Model Overfitting and Underfitting
- Lesson 5: Model Overfitting and Underfitting Prevention
- Lesson 6: Model Combination for Improved Performance
- Section 3: Hyperparameter Tuning
- Lesson 7: Benefits of Hyperparameter Tuning
- Lesson 8: Hyperparameter Tuning Techniques
- Lesson 9: Hyperparameter Tuning Using Amazon SageMaker AMT
- Section 4: Managing Model Size
- Lesson 10: Model Size Factors
- Lesson 11: Model Size Reduction Techniques
- Section 5: Refining Pre-Trained Models
- Lesson 12: Benefits of Fine-Tuning Pre-trained Models
- Lesson 13: Fine-Tuning Pre-trained Models with Custom Datasets on AWS
- Lesson 14: Catastrophic Forgetting Prevention
- Section 6: Model Versioning
- Lesson 15: Benefits of Amazon SageMaker Model Registry
- Lesson 16: Registering and Deploying Models with SageMaker Model Registry
- Section 7: Conclusion
- Lesson 17: Course Summary
- Lesson 18: Assessment
- Lesson 19: Contact Us
