The first step in training a neural network to solve a problem is usually the selection of an architecture: a specification of the number of computational nodes in the network and the connections between them. Architectural decisions are generally based on historical precedent, intuition, and plenty of trial and error.Read More