“The hidden layer is where the (neural) network stores it’s internal abstract representation of the training data.” Ivan Vasilev