Hello everyone, I’m from Rostock/Germany and new here.
In 2014 I created an interactive genetic algorithm for MIDI in pure data to be able to control synthesizers by means of the laws of inheritance established by Gregor Mendel. ( Ewolverine 4 U | PURE DATA forum~ )
It included mutation, recombination, step length adjustment, rules of inheritance and a rudimentary target drive which analyzed synthesized sounds and compared it to a target sample. The results were quite bad, to be honest, as the analysis didn’t account for the time component of audio events very well.
So my question is how I could use an artificial neural network to analyze and compare a synthesized sound to a target sample and let it drive the genetic adaptation by selection of best matches. Any suggestions?