Accession Number : AD0716491

Title :   Generalized Inversion and Automata Morphisms. Part I,

Corporate Author : ECOLE NATIONALE SUPERIEURE DES MINES DE PARIS FONTAINEBLEAU (FRANCE) CENTRE D'AUTOMATIQUE

Personal Author(s) : Depeyrot,M.

Report Date : JUL 1970

Pagination or Media Count : 38

Abstract : Finite-state system theory is based on two main classes of Automata models: transition-output models (T) are most adequate for identification (they include Mealy sequential machines and Shannon communication channels) while state-output models (S) are best fitted for control or decision theoretic studies. An abstract algebraic approach is classically applied to the latter (S) models and can be extended to the former (T) models. A tensor formulation proves to be handy for the study of (stochastic) infinite operands over a finite carrier and of monoid representation morphisms. Here the categorical properties common to both classes of automata are investigated in the stochastic and deterministic cases, showing how spectral analysis is deeply rooted in the new framework. The main tool is that of generalized inverse matrices of affine type, well fitted for the representation of maps and morphisms with a probabilistic meaning. (Author)

Descriptors :   (*ARTIFICIAL INTELLIGENCE, ALGEBRA), (*COMPUTERS, MATHEMATICAL MODELS), AUTOMATA, GROUPS(MATHEMATICS), TENSOR ANALYSIS, MATRICES(MATHEMATICS), STOCHASTIC PROCESSES, DECISION THEORY, ALGORITHMS, FRANCE

Subject Categories : Computer Hardware
      Bionics

Distribution Statement : APPROVED FOR PUBLIC RELEASE