Symmetry

Human artifacts from chess boards to airplanes exhibit symmetries. From the highly regular patterns of circuitry on a microchip, to the interchangeable pistons in a car engine, or seats in a commercial aircraft, we are drawn aesthetically and organizationally to symmetric designs. Part of the appeal of regular or symmetric designs is that they allows us to reason about and understand larger and more complex structures than we could otherwise handle. It follows that if we are to build computer systems that configure, schedule, diagnose, or otherwise reason about human artifacts, we need to endow these reasoning systems with the ability to exploit structure in general and symmetries in particular.

CIRL

CIRL has developed algorithms that detect symmetries in optimization problems, and compile these symmetries into axioms that restrict the search. These "symmetry breaking" axioms are not sound deductions from the constraints in the problem, but do have the property that they neither increase or decrease the optimal value of the objective function.

Pointers

Exploiting Symmetry in Lifted CSPs
Previous work has show how to exploit symmetries in propositional encodings of constraint satisfaction problems. In this paper, problems in more compact lifted (quantified) descriptions are considered. By David Joslin and Amitabha Roy. It appeared in Proc. AAAI in 1997. Compressed Postscript document.
Symmetry Breaking Predicates for Search Problems
Many reasoning and optimization problems exhibit symmetries. In this paper, a general scheme whereby symmetries are exploited by adding "symmetry-breaking" predicates to the theory is presented. The new theory is then solved and experimental results show that in at least two problems, the symmetry-breaking predicates help in solving the problem faster. By James Crawford, Gene Luks, Matt Ginsberg, and Amitabha Roy. It appeared in Proc. KR in 1996. Postscript document.

Parent areas:
Solution structure

Back to the CIRL Overview Page.