knowledge representation using dynamically interlaced hierarchies

(Michalski, Alkharouf, Utz)

This project concerns a development of a new type of knowledge representation that facilitates all kinds of inferences and is thus particularly relevant to the development of multistrategy task-adaptive learning. Dynamic Interlaced Hierarchies (DIH) is based on psychological research into human semantic memory structure and utilizes hierarchies as its basic organizational principle. By storing new knowledge as links between hierarchically organized concepts, a conceptual framework is constructed that can represent very diverse and complex forms of knowledge as well as various knowledge transformations.

DIH uses type and part hierarchies of concepts as background knowledge, or knowledge considered to be relatively stable and unchanging. Statements or facts are stored as links between concepts and are considered dynamic knowledge, as these links are constantly being created and modified, strengthened or weakened. These links have numeric factors (or 'merit parameters') attached that affect the strength of the relationship between the various concepts. Rules and dependencies are bi-directional, each with a separate forward and backward 'strength'.

Inference patterns such as generalization/specialization, abstraction/ concretion, and similarity are easily visualized in DIH. Also these inferences are facilitated, since the procedure consists of manipulating links between hierarchies. Creating new links between concepts represents learning. In this way learning builds upon the background knowledge of the hierarchies and the dynamic knowledge already in place.

Selected References

Alkharouf, N.W. and Michalski, R.S., "Multistrategy Task-Adaptive Learning Using Dynamic Interlaced Hierarchies: A Methodology and Initial Implementation of INTERLACE," Proceedings of the Third International Workshop on Multistrategy Learning (MSL-96), Harpers Ferry, WV, May 23-25, 1996, pp. 117-124.

Hieb, M.R. and Michalski, R.S., "Multitype Inference in Multistrategy Task-adaptive Learning: Dynamic Interlaced Hierarchies," Informatica: An International Journal of Computing and Informatics, Vol. 17, No. 4, pp. 399-412, December 1993.

For more references, seeĀ publications section.


MLI Copyright © 2013 Machine Learning and Inference Laboratory
College of Health and Human Services, George Mason University
4400 University Dr, MSN 1J3, Fairfax, VA 22030, U.S.A