High Performance Knowledge
Bases, DARPA
Sponsoring Agency: AFOSR
Knowledge Acquisition Area
Constructing and Refining Knowledge Bases:
A Collaborative Apprenticeship Multistrategy Learning Approach
HPKB East Coast Meeting
Boston, April 1st, 1997
Gheorghe Tecuci
Learning Agents Laboratory
Computer Science Department
George Mason University, Fairfax, VA 22030
Slide Presentation
- Goals
- KB Development Architecture
- KA Toolkit Architecture
- Main Processes of Knowledge Acquisition
- Illustrations of the KA Approach
- Knowledge Representation: Semantic Network
- Knowledge Representation: Plausible Version Space
Rule
- Acquiring Knowledge to Judge Relevance of Historical
Sources
- Rule Learning
- Explanation of Input Example
- Learning from Explanations
- Generalizing the Explanation to an Analogy Criterion
and Learning a PVS Rule
- Solving Problems by Analogy
- KB Refinement through Active Experimentation
- Learning From a Positive Example
- Generalization of the Plausible Lower Bound
- Learning From a Negative Example
- Specializations of the Two Plausible Bounds
- Extending the Semantic Network as a Result of Rule
Refinement
- Acquiring Knowledge Representing the Expertise of a
Company Commander
- Semantic Network for Company Commander
- Learning a Rule for Platoon Placement
- Learned PVS Rule for Platoon Placement
- KB Refinement
- Specializations of Both Bounds of the Plausible
Version Space
- Correct Placement of Company-d Generated by the KA
Agent
- Updates to the Plausible Lower Bound
- Refined PVS Rule
- Some Research Issues for Rule Learning
- Some Research Issues for KB Refinement
- Conclusions
Go to:
First slide
Back to:
GMU HPKB Project Home Page
HPKB East
Coast Introductory Meeting Briefings
HPKB Home Page
Last Update: 4/23/97