The capacity of a deep learning neural network controls the scope of the types of mapping functions that it can learn. Model capacity can approximate any given function. When there is a higher model capacity, it means that the larger amount of information can be stored in the network.
