Constructing Higher Order Neurons of Increasing Complexity in Cascade Networks

摘要: A problem faced by many constructive neural networks using a cascade architecture is the large network depth. This results in fan-in and propagation delays, problems especially relevant for VLSI implementation of these networks. work explores effect limiting depth cascades created CasPer, algorithm. Instead single hidden neurons, series towers are built. Each tower can be viewed as Higher Order Neuron (HON). The optimal complexity HON required given difficult to estimate, form bias-variance dilemma. overcome via construction HONs with increasing complexity. It shown that constructing this manner chance overfitting reduced, noisy data.