To ensure your machine learning model generalizes well to unseen data and does not suffer from overfitting, you want to evaluate its performance robustly. Instead of relying on a single train-test split, which process partitions the data into multiple subsets, training and testing the model multiple times so that every data point is used for both training and validation?

Cross-validation is a statistical method used to estimate the performance and generalization capability of machine learning models. Cross-validation, such as K-fold cross-validation, involves partitioning the dataset into K equal-sized subsets (folds). The model is trained K times. In each iteration, K-1 folds are used as the training set, and the remaining single fold is used as the validation set for testing. This process repeats so that each of the K folds is used exactly once as the validation data. The results from all K iterations are then averaged to produce a single estimation. This approach ensures that the model evaluation is not biased by a single, arbitrary train-test split and utilizes the entire dataset for both training and validation. Min-max normalization is a data preprocessing technique that scales features to a fixed range, usually 0 to 1, to ensure numerical stability during training. Learning rate decay scheduling is an optimization technique that reduces the step size during gradient descent as training progresses to help the model converge to a global minimum. Hyperparameter grid searching is an optimization method that systematically tests combinations of hyperparameters to find the best configuration, often using cross-validation as its evaluation metric, but it is not the validation process itself. Consequently, cross-validation is the correct term for evaluating model performance across multiple data subsets.