Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Hybrid Artificial Neural System/Expert System

IP.com Disclosure Number: IPCOM000102218D
Original Publication Date: 1990-Nov-01
Included in the Prior Art Database: 2005-Mar-17
Document File: 2 page(s) / 85K

Publishing Venue

IBM

Related People

Land Jr, WH: AUTHOR [+3]

Abstract

Disclosed is a method which utilizes Artificial Neural Systems (ANSs) to handle identification problems where data enter the system in a continuous flow and the characteristics needed for identification vary as a function of time. To evaluate the concept, ANSs were designed and tested which identify Force Elements and communication (COMINT) networks that operate under the control of a rule-based expert system. These Force Elements were identified using COMINT signature inputs, and COMINT NETS were identified using Force Element inputs. The ANSs are monitored and controlled by a Monitoring Expert System (MES) via a global blackboard.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Hybrid Artificial Neural System/Expert System

       Disclosed is a method which utilizes Artificial Neural
Systems (ANSs) to handle identification problems where data enter the
system in a continuous flow and the characteristics needed for
identification vary as a function of time.  To evaluate the concept,
ANSs were designed and tested which identify Force Elements and
communication (COMINT) networks that operate under the control of a
rule-based expert system.  These Force Elements were identified using
COMINT signature inputs, and COMINT NETS were identified using Force
Element inputs.  The ANSs are monitored and controlled by a
Monitoring Expert System (MES) via a global blackboard.

      The figure depicts the implementation of this hybrid system
using commercial hardware and software.  The ANSs and the MES are
separate programs running on dedicated PS/2 Model 80s.

      All communications between he ANS and MES are accomplished by a
simulated global shared memory (blackboard).  The Hybrid ANS/Expert
System is hosted by a distributed processing testbed developed to
host an application called the Distributed Intelligent Defense System
(DIDS) [*] and operates as follows:  The trained ANSs process
buffered COMINT signature data to develop Force Elements and COMINT
networks and write their outputs to the blackboard.  The Force
Element ANSs are trained by signature input patterns coded as a space
of binary n- tuples which the specific Force Elements are expected to
"own" while the training output patterns are unique binary n-tuples
representative of specific Force Elements.  Input/output training
pairs for the COMINT NET ANSs are defined in a similar manner.  The
input components of these binary n-tuples vary as a function of time
in a manner which follows the predicted content of the signature
buffers.  The MES will access ANS outputs and...