Multilayer perceptron design algorithm

This paper describes a design algorithm that has been developed to calculate the number of hidden nodes required and compute a good set of starting weights for the multilayer perceptron (MLP). There are significant advantages to being able to calculate the number of hidden nodes required. The proper choice of the number of hidden nodes results in shorter training times, better generalization, and simpler computations in implementation. This method is then used to design an efficient, effective MLP for multiple-class decision using these simplified binary-decision neural networks. The resulting algorithmic structure has an efficient pipelined implementation. Simulations describe the application of the design algorithm and parametric classification of a transient signal. A modified wavelet feature representation is introduced as an input to the neural networks associated with arrival time discrimination.<<ETX>>