Neural programming, computational scheme and applications.

Title: Neural programming, computational scheme and applications.
Identifier: AAI9405598
identifier: 9405598
Creator: Weiss, Jacob.
Contributor: Adviser: Eralp Akkoyunlu
Date: 1993
Language: English
Publisher: City University of New York.
Subject: Computer Science | Artificial Intelligence
Abstract: An attempt is made to solve difficult AI problems by introducing a new scheme, the net program. The net program is a programmable network of cooperative neural nets and procedural functions. Its goal is to bridge the gap between connectionist-subsymbolic and symbolic processing in AI. A set of subroutines has been devised for "neural programming" as an enhancement to the procedural language C. The resulting language, which we call C{dollar}\sp{lcub}\rm +NET{rcub}{dollar}, enables and controls modular structuring, training and activation of nets and the interaction among nets and procedural functions. We applied C{dollar}\sp{lcub}\rm +NET{rcub}{dollar} on a model for natural language comprehension as a cardinal unsolved AI problem of an intricacy that would strain any conceivable computational scheme and on a model for pattern recognition as a demonstrative example of the advantages of the approach we are presenting. We also present a design of an application to financial forecasting utilizing our net program scheme.
Type: dissertation
Source: PQT Legacy CUNY.xlsx
degree: Ph.D.