Fuzzy Inferencing: A Novel, Massively Parallel Approach

TitleFuzzy Inferencing: A Novel, Massively Parallel Approach
Publication TypeTechnical Report
Year of Publication1995
AuthorsBuller, A.
Other Numbers965
Keywordsdistributed inferencing, fuzzy control, neural networks
Abstract

The report presents a model of a fuzzy control in which decisions are worked out based on results of a competition between groups of agents which, represented by binary words, navigate in a neural working memory. Each agent is endowed with a strategy of its own behavior and carries its opinion. The opinions are symbolic statements encoding facts and/or rules. A Fuzzy Knowledge Base provides rules, as well as the values of membership of given measures to appropriate facts interpreted as fuzzy sets. At a given moment an indoctrinating device generates a fact or its negation with a probability calculated based on related membership value. A debate in the Society of Agents results in a victory of adherents of a particular solution. An ultimate decision is based on a poll. A hardware facilitating this kind of computation, as well as some simulation results are discussed.

URLhttp://www.icsi.berkeley.edu/ftp/global/pub/techreports/1995/tr-95-025.pdf
Bibliographic Notes

ICSI Technical Report TR-95-025

Abbreviated Authors

A. Buller

ICSI Publication Type

Technical Report