Functional Link Networks

A simple transformation of the input data for a feedforward network often dramatically speeds training. The input layer becomes much larger, sometimes impractically so. However, the number of hidden neurons can usually be greatly decreased, or eliminated entirely.