Learning support
Core formula
Expected prediction error = Bias² + Variance + Irreducible error
The best model minimizes the combined error rather than forcing either bias or variance to zero.
Diagnosis guide
High train and validation error: likely underfitting.
Low train, high validation error: likely overfitting.
Low and similar errors: useful generalization.
Common controls
Reduce variance: more data, regularization, bagging, pruning.
Reduce bias: richer features, greater capacity, boosting.
Validate safely: pipelines, cross-validation, untouched test data.
Quiz setup
Choose a mode, topics, difficulty, scoring, and feedback style.
Options
Topics
Question navigator
Keyboard: ← previous, → next, B bookmark, H hint, S save.