Selectionist Musical Automata: integrating explicit instruction and evolutionary algorithms

Peter Beyls


Abstract:

We describe an open, modular system for experimentation with algorithmic composition. It integrates evolutionary methods for implicit synthesis of complex hierarchical structures as well as a large library of software tools allowing user directed explicit object manipulation. The idea is to combine the generative potential of genetic methods with the instructive power of user control embedded in knowledge and culture. Melodic material is generated by viewing the rewrite rules of both L-systems and linear cellular automata as genotypes. The user creates cross-overs and mutations according to visual/auditory interactive evaluation. One can traverse genetic space by morphing one rule into another. Further, objects are visualised in a graphic interface and available for direct manipulation, analysis and structural assembly. A complex harmonisation module creates articulated chords to a given melody based on tension profiles and conventional pattern matching. An embedded interpreter allows the user to extend the functionality of the system which is fully operational and implemented in Object Lisp.

Download paper: [pdf] 533831 bytes