MICA: A Hybrid Method for Corpus-Based Algorithmic Composition of Music Based on Genetic Algorithms, Zipf's Law, and Markov Models
Loading...
Date
2014-01-16
Authors
Nagelberg, Alan
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
An algorithm known as the Musical Imitation and Creativity Algorithm (MICA) that composes stylistic music based on a corpus of works in a given style is presented. The corpus works are digital music scores created from the widely available MIDI format. The algorithm restricts the note placement in compositions using a Markov chain model built from discrete-time representations of the corpus pieces. New compositions are evolved using a genetic algorithm with a fitness function based on Zipf's Law properties of various musical metrics in the corpus pieces. The resulting compositions are evaluated by a panel of both musical and non-musical volunteers in a blind survey.
Description
Keywords
Artificial, Intelligence, Computational, Creativity, Music, Machine, Learning, Algorithmic, Composition