Postby chung » Oct 14, 2019 19:33
(14/10/2019) random neural network option added
i noticed that spectrum harmonics correction has a structure of a single hidden layer neural network so i tried to generalize it , with a randomize optimization algorythm, seems to work better or is it just an illusion ?
neural network like self training midifile feedback works well (use the output approximated midifile as input for training with the current wav file itself )