Machine Learning Fundamentals · free preview
Capacity, Underfitting and Overfitting
A model class that can fit arbitrary labels cannot certify that it found a pattern, and the shapes of the training and validation errors diagnose whether capacity is too small or too large.
1. What low training error certifies
What low training error certifies
2. Reading the two errors
Reading the two errors
3. The effect of more data
The effect of more data
4. Which regime the model is in
A classifier's training error is 2% and its validation error is 31%. Which statement describes this regime and the effective response?
5. The underfitting signature
A model's training error is 34% and its validation error is 36%, with a small gap between them. This pattern points to ____.
Practice this interactively
Free account · instant grading · spaced review that schedules itself.
Start this course for free