- Feedback and Q&A, especially on Homework 2
- Simplifications to the Reflections process
- Continuing intro to AI Tech Questions: slides
- How to evaluate a classifier? Accuracy and confusion matrices
- time permitting loss functions and the importance of partial credit
Examples:
- If the classifier was actually just looking at the first letter of the image filename, it could get 100% accuracy but it wouldn’t actually be learning to recognize images.