Representing High-Level Knowledge Structures in Massively Parallel Networks
Chun, Hon Wai
This item is only available for download by members of the University of Illinois community. Students, faculty, and staff at the U of I may log in with your NetID and password to view the item. If you are trying to access an Illinois-restricted dissertation or thesis, you can request a copy through your library's Inter-Library Loan office or purchase a copy directly from ProQuest.
Permalink
https://hdl.handle.net/2142/69368
Description
Title
Representing High-Level Knowledge Structures in Massively Parallel Networks
Author(s)
Chun, Hon Wai
Issue Date
1987
Doctoral Committee Chair(s)
Waltz, David L.
Department of Study
Electrical Engineering
Discipline
Electrical Engineering
Degree Granting Institution
University of Illinois at Urbana-Champaign
Degree Name
Ph.D.
Degree Level
Dissertation
Keyword(s)
Artificial Intelligence
Computer Science
Abstract
Traditional artificial intelligence (AI) research has concentrated mostly on modeling high-level thought processes such as problem-solving and planning, and high-level representations such as rule-based heuristics and frame-based knowledge structures. Massively parallel networks, on the other hand, have been used mainly to model low-level perceptual processes such as vision, speech, associative memory, and learning. Recent research has started to bridge the gap between these disciplines. Massively parallel networks have many representational and computational advantages to bring to traditional AI work. These networks are very good at filling in partial information and at learning and representing subtle relationships among concepts. In addition, they provide a means to tightly integrate information from different sources as well as a model to encode parallel processing. However, many difficult issues need to be solved before massively parallel techniques can become more applicable and be able to complement traditional AI techniques. These issues include the problem of variable binding, multiple instantiations of knowledge structures, recursion, hierarchical abstraction, and temporal constraints.
These problems are explored in this thesis through the presentation of massively parallel models for schema structures, an approach to defining knowledge modules, and a model to represent rules. The massively parallel representational techniques introduced in this thesis include a representation of constraints which has been used to model schemata with temporal sequence constraints, a representation of modular knowledge structures that have been used to model adaptive competition as well as interactions among higher-order concepts, a hybrid schema representation which combines connectionist value-passing with marker passing, and a hybrid natural language rule-based system. These techniques have been implemented and successfully tested in applications such as speech recognition and natural language understanding.
Use this login method if you
don't
have an
@illinois.edu
email address.
(Oops, I do have one)
IDEALS migrated to a new platform on June 23, 2022. If you created
your account prior to this date, you will have to reset your password
using the forgot-password link below.