Bias variance overfitting
Explain bias vs variance and how you detect overfitting in a supervised ML model.
Answers use simple, clear English.
Audio N/AQuick interview answer
Bias is error from overly simple assumptions (underfit); variance is sensitivity to training noise (overfit). Overfitting shows as low train error but high validation/test error and a widening train–val gap on learning curves. Fixes: more data, regularization, simpler model, dropout/early stopping, better CV.
Detailed answer
Bias is error from overly simple assumptions (underfit); variance is sensitivity to training noise (overfit). Overfitting shows as low train error but high validation/test error and a widening train–val gap on learning curves. Fixes: more data, regularization, simpler model, dropout/early stopping, better CV. Never tune on the final test set.
Real example & use case
Decision tree depth 40 perfect on train, poor on val → prune or use random forest.
Pros & cons
Pros: actionable diagnosis. Cons: need clean validation design to trust the gap.