← AI & ML
Train it, then test it
See it
Don't grade on the practice questions
- Imagine studying for an exam with the exact exam questions. You'd ace it but learn nothing.
- So we split data in two: training data (to learn) and test data (to check).
- The model never sees the test data while learning.
Training
learn from this
Test
hidden until the end
Score
how well did it do?