Concept
Overfitting
A model learning noise in the training data, so it does well in training and badly on new data.
What should you learn before Overfitting?
- Feature engineeringTurning raw data into inputs a model can learn from: scaling, encoding and new features.
- Machine learningBuilding systems that learn patterns from data instead of following hand-written rules.
- Supervised learningLearning from labelled examples to predict a label or value.
- Model evaluation and metricsYou spot overfitting by comparing training and test scores.
What to learn after Overfitting
- RegularizationPenalising model complexity (L1, L2, dropout) to reduce overfitting.
Connections to other concepts
- Reduced by
- Regularization
- Related
- Bias–variance trade-off
Practise Overfitting
- Machine Learning Interview Questions for Freshers2 questions on Overfitting