An algebraic approach to the autonomously self-adaptable Boolean neural nets (Q2763524)

The Boolean Neural Nets (BNN) were first proposed by McCulloch-Pitts as tool to compute a logical calculus. Later on Caianiello introduced an effective algebraic formalism to represent Boolean neural net. This monograph presents an algebraic theory of the autonomously self-adaptable BNN. Ch. 1, The Boolean neural net, is devoted to the McCulloch-Pitts and Caianiello net formalisms. As illustrative examples, the BNNs computing the propositional calculus are presented. Ch. 2, The knowledge representation in the BNN, discusses the knowledge representation within a system like a Turing Machine black box, a General Purpose Architecture, a Central Nervous System of a vertebrate, and BNN. Ch. 3, The data structure embedded in a BNN, proposes a new way to store the data in some traditional data structures actually exploiting the redundancy present in the data when represented as symbol sequences. Ch. 4, The BNN and the Hebbian rule, implements a Hebbian rule (or Hebbian learning) in BNN to obtain the autonomous trainable nets. In Ch. 5, The adaptable BNN, or ABNN, a necessary condition is given to implement the Hebbian rules in a BNN, and some sufficient conditions for the wiring diagrams guaranteeing a properly working ABNN are discussed. Ch. 6, A GABNN as an autonomously self-trainable control system, discusses a General ABNN as the adaptable control system for writing robot and presents the training procedures for general assemblies. In Ch. 7, The Java package it.na.cy.nnet, the package for the development of neural network simulators is presented. The text is self-contained but assumes experience with the basic concepts and techniques of artificial intelligence.