Writing Good Logic Models

John Hooker

 

Abstract:

Logic-based modeling can result in problem formulations that are natural

and easy to debug. To ease solution, they should be written or

automatically reformulated to achieve two goals: to be as nearly

consistent as possible (in the constraint programming sense) and to gave

a tight continuous relaxation. The former reduces search and

backtracking, and the latter provides better relaxation bounds. This

talk shows how to accomplish these goals for formulas of propositional

logic, cardinality clauses, and implications between cardinality

clauses.