PURPOSE: To efficiently and flexibly remove restrictions by realizing plural wide-range restrictive formulas on a neural network to perform parallel execution.
CONSTITUTION: This system is characterized by not only expressing variables and constants in plural restrictive formulas given to a variable manager 13 by nodes on the neural network where plural nodes 11, which can have states set from the external and subject the sum signal due to addition of plural input signals to nonlinear processing and output the results, are coupled in both directions by links 12 which multiply the signals outputted from one nodes by weights set from the external and output the results to the other nodes but also expressing plural restrictive formulas on the neural network by interpretation of degrees of importance of addition, subtraction, multiplication, and Boolean operation included in plural given restrictive formulas or respective restrictive formulas in a restriction interpreter and setting of weights of links between nodes expressing the variables and the constants.
| JPH03102459 | NEURAL CIRCUIT NETWORK DEVICE |
| JPH0620074 | ELECTRONIC ELEMENT FOR NEUROCOMPUTER |
| JPH076146 | PARALLEL DATA PROCESSING SYSTEM |
SEO KAZUO
ARITA HIDEKAZU
KOBUNE RYUICHI
YOKOTA TAKASHI
ABE KAZUHIRO
Next Patent: SYSTEM FOR AUTOMATICALLY DETERMINING COMBINATION RELATIONS BETWEEN SIMPLE SENTENCES AT THE TIME OF O...