8 Overfitting and Generalization Free Online FlashCards

What is generalization in machine learning?

Generalization is the ability to make accurate predictions on previously unseen data.

What is overfitting?

Overfitting occurs when a model learns training data too closely, including noise or accidental details, and performs worse on new data.

What is underfitting?

Underfitting occurs when a model fails to capture important patterns and performs poorly even on its training data.

What performance pattern suggests overfitting?

Good training performance paired with much worse validation performance suggests overfitting.

What performance pattern suggests underfitting?

Poor performance on both training and validation data suggests underfitting.

Why can excessive model flexibility cause overfitting?

Overfitting is more likely when a model is too flexible for the amount or quality of available data.

How can you reduce overfitting?

Reduce effective complexity: for example, limit tree depth, use fewer relevant features, apply regularization, or stop training early when validation performance stops improving.

How can you address underfitting?

Use a more suitable model or features, reduce excessive regularization, or train longer if training has not converged.

How can more representative training data help generalization?

More varied, representative training examples can help a model learn stable relationships instead of memorizing a small sample.

What are the roles of training, validation, and test data?

Train on training data, use validation data to compare models or tune settings, and reserve test data for final evaluation.

Why should you avoid repeatedly tuning against the test set?

Repeatedly choosing settings based on the test score can make that score overly optimistic, so keep the test set for final evaluation.

How should you prevent leakage during data preparation?

Split the data before preprocessing or feature selection so information from validation or test data cannot leak into training.