Gene expression synthesis

Gene expression programming presents an alternative approach in the evolutionary computation paradigm evolving populations of candidate solutions as valid computer programs that can be used for a potentially wide range of problem solving tasks, including sound synthesis. This paper proposes Gene Expression Synthesis (GES) as a method to evolve sound synthesis functions as nested graphs of unit generators. The functions are encoded into linear chromosomes according to the principles of gene expression programming and then evolved by subjecting the functions to genetic operations and evaluating fitness. The design of the fitness functions involves statistical methods and machine listening algorithms in an attempt to automate the supervision of the synthesis process. The specification of synthesis parameters explores the idea of artificial co-evolution. A parallel population of functions share their fitness values with their respective synthesis functions for which they compute parameter values, while being subjected to genetic operations including recombination separately.