| | |
Summary: PDPAR 2004
Non-Clausal Reasoning
Toby Walsh
SFI Research Professor, Cork Constraint Computation Centre, University College
Cork, Cork, Ireland, tw@4c.ucc.ie
Abstract
State of the art SAT solvers typically solve SAT theories encoded into CNF using
DPLL based algorithms. Most problems, however, are not originally expressed
in CNF but contain arbitrary propositional formulae. The original problem must
therefore be converted into CNF. Converting to a simple and uniform structure
like CNF provides conceptual and implementational simplicity. Indeed, a number
of key techniques that improve the effectiveness and efficiency of DPLL solvers
exploit the simple clause structure. However, converting to CNF loses a considerable
amount of information about the problem's structure. This is information that
could be used to improve the search efficiency. I will discuss how a SAT solver
can be implemented to reasoning efficiently with non-clausal formulae, and where
computational advantages can be gained.
This is a joint work with Fahiem Bacchus and Christian Thiffault.
|
