Title :
A new digital neural network and its application
Author :
Tuazon, J.O. ; Hamidian, K. ; Guyette, L.
Author_Institution :
Dept. of Electr. Eng., California State Univ., Fullerton, CA, USA
Abstract :
A new digital model of the neuron, called K/N gate, is presented. The K/N neural networks have important potential application in such areas as pattern recognition and image processing. The proposed model is applied to pattern recognition, like the decimal digits in presence of random noise, and comparison is made with its analog counter part perception model. The results show that the K/N net performs better than the perceptron net. The advantages of the proposed model over the analog model are reported as follows: 1) no time consuming training is needed; 2) the desired pattern can be easily distinguished from possible noise inputs given the fact that the distance between the two has been predetermined and included in the design; and 3) the implementation can be achieved with memory and register architecture
Keywords :
digital integrated circuits; logic devices; logic gates; neural chips; pattern recognition; random noise; K/N gate; digital neural network; neuron model; pattern recognition; random noise; register architecture; Biological neural networks; Biological system modeling; Counting circuits; Image recognition; Nervous system; Neural networks; Neurons; Operational amplifiers; Pattern recognition; Supervised learning;
Conference_Titel :
Electrical and Computer Engineering, 1993. Canadian Conference on
Conference_Location :
Vancouver, BC
Print_ISBN :
0-7803-2416-1
DOI :
10.1109/CCECE.1993.332189