INFERENCE RULE DETERMINATION METHOD AND INFERENCE APPARATUS.

摘要

The inference rule determination method of the present invention is a method which, when input data is given, determines sequentially a membership function representing how much the condition of the IF section of each inference rule is satisfied, by use of a learning function of a nerve circuit network model and obtains an optimum inference result without using an empirical rule. The inference apparatus of the present invention comprises a membership value determination unit (1) which uses ''IF ... THEN ... '' type inference rule, embraces all the IF sections and constitutes a nerve circuit network, individual inference quantity determination units (21) - (2r) which correspond to the THEN sections of each inference rule and determine an inference quantity for each inference rule and a final inference determination unit which judges them collectively to determine the final inference result. If an individual estimation quantity determination unit (2) has a nerve circuit network structure, an inference result can be obtained with a high level of inference accuracy by utilizing the non-linearity of a nerve circuit network model even when an object of inference is non-linear.