Ageneticalgorithmselectedcombinationsofattributesfor amachinelearningsystem.Thealgorithmused90Bach choralemelodiestotrainmodelsandrandomlyselectedsets of10choralesforevaluation.Compressionofpitchwas usedasthefitnessevaluationcriterion.Thebestmodels wereusedtocompressadifferenttestsetofchoralesand theirperformancecomparedtohumangeneratedmodels. G.A.modelsoutperformedthehumanmodels,improving compressionby10percent.
Mark A. Hall