Practical Neural Network Recipies in C++

44

Chapter 3

Data Compression

Three-layer feedforward networks trainedfor autoassociative recall are

sometimes used for data compression. These networks are discussed

in detail starting on page 77, so their inner workings will be glossed

over here. All that is important to understand at this point is that

this model consists of the usual hypothetical input layer and the

actual output layer, with a single layer of neurons hidden between the

input and output

layers.

In general, there are far fewer neurons in the

hidden layer than in the input and output layers. A three-layer

feedforward network having six input and output neurons and two

hidden-laye ...