| Summary: | In any evolutionary search system, the fitness raters are most crucial in determining successful evolution. In this paper, we propose a Hierarchical Self Organizing Map based sequence predictor as a fitness evaluator for a music evolution system. The hierarchical organization of information in the HSOM allows prediction to be performed with multiple levels of contextual information. Here, we detail the design and implementation of such a HSOM system. From the experimental setup, we show that the HSOM's prediction performance exceeds that of a Markov prediction system when using randomly generated and musical phrases.
|
|---|