How can overfitting be reduced?

• A. Increase model complexity
• B. Cross-validation, regularization, more data
• C. Ignore validation
• D. Use fewer samples

Answer: (B)

Overfitting happens when the model learns noise in the training data rather than the real patterns, so the goal is to improve generalization. Using cross-validation gives a more honest estimate of how the model will perform on new data, helping detect and prevent overfitting. Regularization adds a penalty for large weights, effectively keeping the model simpler and less likely to fit random noise. Providing more data, or augmenting existing data, exposes the model to a broader range of examples and reduces the chance it fixes on idiosyncrasies of a small dataset.

Increasing model complexity tends to make overfitting worse because the model can fit more random fluctuations in the training data. Ignoring validation removes the feedback you need to notice when the model isn’t generalizing well. Using fewer samples deprives the model of information and can amplify noise, leading to poorer generalization.