Complex Architecture

Magnus is working on a machine-learning model for his thesis. He has been building a deep neural network that uses a very complex architecture.

The problem is that his model is not generalizing well. Despite having a low error rate on the training set, the error rate on the validation set is much higher.

Magnus knows that his model is suffering from overfitting. He needs to find a way to improve his model's performance on the validation set.

Which of the following techniques will help Magnus fix his model's overfitting problem?

Magnus should try using a smaller network architecture.

Magnus should add more training data.

Magnus should increase the learning rate.

Magnus should add L1 and L2 regularization to the cost function.