Practical Neural Network Recipies in C++

10

Designing Feedforward

Network Architectures

Certain problems are more easily solved by particular network

architectures. For multiple-layer feedforward networks, the number

of hidden neurons can mean the difference between success and

failure. While there are no hard formulas known at this time, this

chapter briefly discusses some rules of thumb for choosing an

appropriate network model and defining specific characteristics of that

model.