Description
Theo
Plan-then-compile architectures attempt to integrate planning, reactivity, and knowledge compilation learning, in order to improve system reactivity while retaining system flexibility. (The Entropy Reduction Engine operates in a similar manner, but is not considered a Plan-then-compile architecture.)
Theo is an attempt at creating a self-improving problem solver that employs generalization mechanisms to improve its problems solving performance at various tasks. In general, self-improving systems must address several issues:
- how to form general concepts from examples
- which concepts to learn
- when to learn
- from what data and knowledge to learn
- how to index what it learns
Theo is a framework for solving these issues, involving problem solving, learning and representation. Its methodologies draw from previous attempts at this scheme, including Soar [Laird, 1987], Eurisko [Lenat, 1983], and PRODIGY [Minton, 1987].
Theo is intended to support:
- Basic research on general problem solving, learning, and knowledge representation, and the interactions between these.
- Accretion of research results in a form that others may reuse and extend.
- An efficient framework for developing effective knowledge-based systems.
Press HERE for a map of the architecture.
To return, press HOME.
To go to the next document, press NEXT.