FAWCETT@RED.RUTGERS.EDU (Tom Fawcett) (07/02/86)
DISCOVERY OF ALGORITHMS
FROM WEAK METHODS
Armand E. Prieditis
Weak problem-solving methods (e.g. means-ends analysis, breadth-
first search, best-first search) all involve a search for some sequence
of operators that will lead from an initial state to a goal state.
This paper shows how it is possible to learn operators whose bodies
contain algorithmic control constructs (e.g. loops, sequences,
conditionals) such that the control construct itself applies the
sequence needed to lead from the initial state to a goal state without
a search for the sequence. By using explanation-based generalization
[EBG] and an explicit theory of algorithms, the method learns
operators (whose bodies contain algorithmic control constructs) that
represent logically valid generalizations of the solution sequence.
Where: Hill Center, Room 423
When: Tuesday, July 15th
Speaker's EMail address: PRIEDITIS@RED.RUTGERS.EDU