Knowledge MaxxingCross-validation
Cross-validation repeatedly splits data into different training and validation portions, evaluating a model on each split and averaging the results for a more reliable estimate.
Prerequisite: Validation set
What matters
Averaging across multiple splits reduces the risk that one lucky or unlucky single split makes a model look better or worse than it really is.
Example
5-fold cross-validation trains and evaluates a model five separate times, each time using a different fifth of the data as validation, then averages the five scores.
Common misconception
Cross-validation does not replace the need for a separate test set; it is typically used during development, with the test set still held out for a final honest check.
Recall question
Why does averaging across several splits give a more reliable estimate than one split alone?