Bias, Variance, Overfitting and Regularisation

Why models underfit or overfit, the bias–variance decomposition, learning curves, cross-validation, and L1/L2, dropout and early stopping.