Related Experiment Videos
One framework, two systems: flexible abductive methods in the problem-space paradigm applied to antibody
J W Smith1, A Bayazitoglu, T R Johnson
1Division of Medical Informatics, Ohio State University, Columbus 43210, USA.
Artificial Intelligence in Medicine
|June 1, 1995
Summary
This study introduces flexible knowledge-based systems using a problem-space architecture for opportunistic problem-solving. These systems adapt methods based on available knowledge, ensuring robust performance in diverse tasks.
Area of Science:
- Artificial Intelligence
- Knowledge Representation and Reasoning
- Cognitive Science
Background:
- Traditional knowledge-based systems often require rigid method and task representations.
- Adapting problem-solving methods to specific situations is a key challenge in artificial intelligence.
Purpose of the Study:
- To develop flexible knowledge-based systems capable of opportunistic adaptation of problem-solving methods.
- To create a general and robust abductive problem-solving method adaptable to varying knowledge.
Main Methods:
- Utilized a problem-space architecture for opportunistic method adaptation.
- Developed a flexible abductive problem-solving method based on subgoals and preferences.
- Implemented domain-specific knowledge to override or augment the basic method.
Main Results:
- Demonstrated that additional knowledge significantly alters system behavior at runtime.
- Showcased systems (RedSoar, LiverSoar) that perform problem-solving even with limited domain knowledge.
- Achieved robust and non-brittle problem-solving through a general architecture and flexible method.
Conclusions:
- Flexible knowledge-based systems can be effectively built using opportunistic adaptation.
- The proposed abductive problem-solving method offers generality and robustness across domains.
- Runtime knowledge dynamically shapes problem-solving behavior, enhancing system adaptability.