Module 3: Evaluating and Optimizing Transfer Learning

Evaluating and Optimizing Transfer Learning banner

Topics: The following topics are covered in this module:

Choosing the Right Pre-Trained Model
Transfer Learning vs. Full Training
Evaluating Transfer Learning Techniques
Performance Evaluation: Measuring the Effectiveness of Transfer Learning
Common Transfer Learning Issues & Troubleshooting Strategies

Objectives

By the end of this module, students will be able to:

Critique a pre-trained model’s suitability for use in transfer learning to a new task and select effective candidate models.
Assess the trade-offs of transfer learning versus full training and make justifiable choices for their approach.
Develop intuition on which, if any, transfer learning techniques to use in model development.
Evaluate the outcome of transfer learning techniques (feature extraction, fine-tuning and LoRA) on model performance.
Evaluate the outcome of transfer learning techniques compared to a similar model trained without transfer.
Troubleshoot and remediate common transfer learning model issues like catastrophic forgetting and negative transfer.

Watch

Video: Optimizing and Evaluating Transfer Learning (11:55)

Topics

Click on the accordions links read the materials within.

Practice and Apply

Pre-Trained Model Evaluation Exercise

In this last exercise, we’ll compare different pre-trained models for their suitability as the Source model for feature extraction. Then we’ll practice using evaluation metrics to guide us in improving the new model. Happy coding!