« A Method of Optimization using Production Rules and Local Evaluation Functions and the Macroscopic Behavior during its Computation Processes | Main | A Method of Solving Constraint Satisfaction Problems using Production Rules and Local Evaluation Functions »

Symbol Processing as Stochastic Processes - Macroscopic Models of Computation Processes --

Kanada, Y., Technical Report of IEICE, COMP92-93 and SS92-40, The Institute of Electronics, Information and Communication Engineers, pp. 1-10, 1993, Published by IEICE.

[ 日本語のページ ]
[ Paper PDF file (in Japanese) ] [ Paper postscript file (in Japanese) ]
[ OHP postscript file: Slides, Handout ]
[ OHP PDF file: Slides, Handout ]

Abstract: Systems, which perform complex symbol processing, can neigher be understood by understanding the program statically nor by tracking their microscopic behavior. Thus, macroscopic models of the computation processes are necessary to understand the systems. Macroscopic models are also necessary for automatic control of computation and realizing self-organizing computation. This paper examines foundation of macroscopic theory of computation processes, and proposes modeling based on stochastic process theory. This paper also gives an example of a macroscopic model, the Markov chain model of the Chemical Casting Model (CCM), which is a microscopic model, and analizes computation processes of a graph coloring problem based on CCM. This theory realizes a fusion of symbol computation in the microscopic model and pattern computation in the macroscopic model.

Introduction to this research theme: CCM: Chemical-Computation Model

Keywords: CCM, Symbol Processing, Stochastic Process, Emergent computation, Randomized computation, Randomized problem solving, Rule-based computation, Rule-based problem solving, Stochastic Process, Local information, Localized computation, Self organization, Micro model, Macro model, Symbolcomputation, Pattern computation

Post a comment


This page contains a single entry from the blog posted on March 1, 1993 12:00 AM.

Many more can be found on the main index page or by looking through the archives.

(C) Copyright 2007 by Yasusi Kanada
Powered by
Movable Type 3.36