Machine learning in practice

Fit preprocessing without leakage, interpret metrics, implement neighbours and regression, score splits and stop training.

Progress and your code are kept in this browser only.

1. 1Fit preprocessing on training data only
2. 2Read a confusion matrix before accuracy
3. 3Keep binary log loss finite
4. 4Classify with nearest neighbours
5. 5Differentiate a regression loss by hand
6. 6Choose a split by weighted impurity
7. 7Build disjoint evaluation folds
8. 8Track improvements with patience