Genetic Strategy Selection for Parallel Theorem Proving

Automated theorem provers use search strategies for accomplishing their proof tasks. Unfortunately, there is no unique strategy which is uniformly successful and uniformly fast on all problems. The combination of more than one strategy in parallel leads to a decreasing response time compared with sequential approaches and to an increasing number of solutions found. Limited availability of resources as time or processors necessitates the efficient use of these resources by partitioning the resources adequately among the involved strategies. We describe the paradigm of strategy parallelism for general search problems, and sketch some problems which have to be solved when constructing and configuring a strategy parallel prover. Then we present a solution for this prover configuration problem which works quickly, with high precision, and fully automatically. We finish giving some experimental data for our implementation of this concept, that has been proven to be one of the fastest availab..