Untouched issues
Feature engineering
- From measurements to features
Feature selection
- Discovering the relevant features
Smoothing and priors
- Not enough data to compute robust statistics
Representing selection bias
- All the subjects from the same population