Accurate Analysis of Combinatorial Problems: From the Particular to the General

Combinatorial problems pervade all aspects of our social, environmental and economic life, but finding good
solutions to these problems can take too much computer time. While there are several techniques that can
considerably reduce this time, the speedups are only guaranteed if the combinatorial problems have certain
formal properties. Detecting whether such properties occur has proved difficult: approaches that take the
problem data into account are very accurate but too costly; those that do not, are efficient but inaccurate. We
investigate ways of combining these two approaches into a general framework capable of accurately inferring
useful properties of combinatorial problems at low cost.