Overfitting & Underfitting
The two ways a model fails — and how to fix each.
- ✓What overfitting and underfitting look like
- ✓How to diagnose each
- ✓Fixes like more data and regularisation
Explanation
Overfitting is memorising the training data, including its noise — great train score, poor test score. Underfitting is being too simple to capture the pattern — poor on both.
Diagnose by comparing train vs test performance: a big gap means overfitting; both low means underfitting.
Fixes for overfitting: more data, a simpler model, or regularisation (penalising complexity). Fix for underfitting: a more capable model or better features.
A student who memorises past exam answers (overfit) aces practice papers but struggles with new questions — exactly how overfitting behaves.
- • Judging a model only by training accuracy and missing that it overfit.
Explain overfitting vs underfitting using a real-life analogy of your own.
1. A model scores 99% on train but 65% on test. This is...
2. Which helps reduce overfitting?