Evolving Compression Preprocessors With Genetic Programming

We present a new approach for applying genetic programming to lossless data compression. Unlike programmatic compression the evolved programs are preprocessors. These preprocessors aim at enhancing the compression rate of the given data by transforming it. The entropy based tness function is both fast and independent of the type of information being processed. The obtained results are encouraging in sense that signi cant improvements can be achieved. Furthermore the required computation time is much smaller than in the case of programmatic compression, making the presented approach more viable. We used a strongly typed GP kernel. The kernel o ers the extra advantage of being able to exploit parallel execution through the island model.