Memory Efficient Evaluations of Nonlinear Stochastic Equations and C3 Applications.
Abstract
The Statistical Mechanical Neural Computer (SMNC) developed in this thesis utilizes a Statistical Mechanical Nonlinear Algorithm (SMNA) to determine the long-time probability distribution of highly nonlinear stochastic systems. The use of the SMNA and a novel mesoscopic scaling technique help provide the SMNC with the capabilities of neural computers without the drawbacks of huge connection matrices and their attendant computational requirements. In this thesis, the SMNC is initially used to verify the ability of the SMNA to duplicate relatively simple, single variable path integral solutions to nonlinear Fokker-Planck equations. After the fundamental algorithms are validated, the SMNC's ability to simulate a two-variable, multicellular problem by modeling a portion of the neocortex consisting of 100,000 neural units is discussed. There are many important applications of the SMNC and its unique SMNA to C3 systems including radar, sonar and electronic signals processing, missile guidance systems and an integrated battle management system. Such C3 systems will benefit from the SMNC's potential to efficiently filter large amounts of data, recognize patterns and anticipate, with some degree of uncertainty, the future state of highly nonlinear stochastic systems. Keywords: Command, Control and Communications, Monte Carlo simulation; Nonlinear probability distributions; Neural computers; Computer simulation; Modeling.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1987
- Accession Number
- ADA189872
Entities
People
- John C. Connell Jr.
Organizations
- Naval Postgraduate School