Regularization in machine learning reduces overfitting by penalizing model complexity. L1 and L2 are common methods that improve generalization on unseen data