Validation Set
A specific subset of data held back during model training, used to iteratively evaluate performance, tune hyperparameters, and detect overfitting.
Think of It Like This
Like a student taking a series of practice tests to gauge their readiness, entirely separate from their study notes and the final exam.
While the model actively learns from the training set, the validation set is untouched by gradient descent. It acts as a surrogate for unseen real-world data. If training loss decreases but validation loss increases, the model is overfitting, signaling that early stopping or heavier regularization should be immediately applied.