Regularization always gives worse performance on the training set, but the benefit is (when properly optimized) it gives better performance on the validation set.
thanks
could you give me intuition why does it give worse performance on the training set and give better performance on the validation set ??