Abstract

Higher-order neural networks¹have several desirable computational characteristics such as significantly decreased convergence times, increased storage capacity over first-order layered networks (which respond only to weighted sums of the inputs ${\displaystyle \sum W_i{x}_i}$), and an ability to encode a prioriknowledge in the network. Simulation results for the problem of learning to classify clusters in a binary string showed that a second-order network (which also includes weighted sums of products ${\displaystyle \sum W_{ij}{x}_i{x}_j}$) generalized correctly 90% of the time after training on only 10% of the training set.

© 1987 Optical Society of America