1 March 1991 Influence of different nonlinearity functions on perceptron performance
Author Affiliations +
Proceedings Volume 1396, Applications of Optical Engineering: Proceedings of OE/Midwest '90; (1991) https://doi.org/10.1117/12.25814
Event: Applications of Optical Engineering: Proceedings of OE/Midwest '90, 1990, Rosemont, IL, United States
Abstract
Influence of two new nonlinearity functions on Perceptron performance is studied. The two new functions under consideration are Gaussian and sinusoid functions. The new functions create multithreshold Perceptions capable of handling both binary and analog inputs. A computer program has been developed to simulate behavior of a network utilizing either of the two modified Perceptrons. Both XOR and Parity Check problems were solved using a single-layer network utilizing these modified Perceptions. Based on the results obtained from the simulation the modified Perceptions are capable of solving problems (such as XOR) that can not be solved using a single-layer of the classical Perceptron. Also networks utilizing these modified Perceptions require fewer number of iterations to converge to a solution than that of a multi-layer network of classical Perceptions using back propagation. In addition the results show that Sinusoidal Perceptronperforms better than Gaussian Perception. 1.
© (1991) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Ashenayi Kaveh, Ashenayi Kaveh, James Vogh, James Vogh, } "Influence of different nonlinearity functions on perceptron performance", Proc. SPIE 1396, Applications of Optical Engineering: Proceedings of OE/Midwest '90, (1 March 1991); doi: 10.1117/12.25814; https://doi.org/10.1117/12.25814
PROCEEDINGS
11 PAGES


SHARE
RELATED CONTENT


Back to Top