Capturing the dynamics of cellular automata, for the generation of synthetic persian music, using conditional restricted Boltzmann machines

Abstract

In this paper the generative and feature extracting powers of the family of Boltzmann Machines are employed in an algorithmic music composition system. Liquid Persian Music (LPM) system is an audio generator using cellular automata progressions as a creative core source. LPM provides an infrastructure for creating novel Dastgāh-like Persian music. Pattern matching rules extract features from the cellular automata sequences and populate the parameters of a Persian musical instrument synthesizer [1]. Applying restricted Boltzmann machines, and conditional restricted Boltzmann machines as two family members of Boltzmann machines provide new ways for interpreting the patterns emanating from the cellular automata. Conditional restricted Boltzmann machines are particularly employed for capturing the dynamics of cellular automata.