Pairwise Sequence Comparison and the Genetic Programming of Iterative Concurrent Programs

The genetic programming of iterative concurrent programs written in the CCS process algebra is investigated. Using a generational genetic programming scheme, experiments succesfully evolved a cyclic concurrent program that performs a even-parity-2 analysis on a communicating input stream. The fitness evaluation strategy determines how well programs communicate with randomly generated streams of input signals. Fitness is measured by performing a pairwise sequence alignment comparison of two execution sequences – the output sequence generated by the program communicating with the test signal stream, and the correct output sequence for that test case. The optimal edit distance between these sequences is efficiently computed using dynamic programming. The main result is that sequence alignment evaluation against randomly generated test cases is a promising evaluation strategy for evolving cycling protocols.