Add to BibSonomyAbstract
Use of the linear discriminant function in the first-order neural network (e.g., the Hopfield model) has drawbacks in that the first-order correlation signal is easily interfered by the low-correlated random noise and this kind of neural network has limited storage capacity. High-order neural networks that can effectively suppress the noise with a high-order nonlinear discriminant function in the correlation domain were recently proposed.
© 1989 Optical Society of America
PDF ArticleMore Like This
Optical quadratic neural networks
A. P. Ittycheriah, John F. Walkup, and Thomas F. Krile
THZ3 OSA Annual Meeting (FIO) 1988
Learning in Interpolation Networks for Irregular Sampling: Some Convergence Properties
Albert J. Ahumada and Jeffrey B. Mulligan
WC3 Applied Vision (AV) 1989
Near neighbor neurons interconnected neural network
Shaoping Bian, Kebin Xu, and Jing Hong
THT27 OSA Annual Meeting (FIO) 1989