Subject: Re: [linux-audio-user] EVSynth - genetic algorithms and sound synthesis
From: Steve Harris (S.W.Harris_AT_ecs.soton.ac.uk)
Date: Wed Feb 20 2002 - 17:46:19 EET
On Wed, Feb 20, 2002 at 03:15:14 +0000, Dave Griffiths wrote:
> > What are you using as your training critera? Human feedback? Or do you
> > have "goodness" metric?
>
> It's old fashioned human choice for now. It would be possible with the L
> System to have selection criteria like complexity, pitch range,
> repetition etc. They could be mapped to sliders, and have the thing
> evolving the whole time, morphing between generations.
>
> It's just a question of working out the algorithms that gauge those
> selection parameters really.
Yep, thats hard.
One I considered a few years back is similarity: have an input sound at
different pitches and the pitch and the metric is the similarity of the
output sound and input sound.
That way you could have a kind of learning synth. I'm not sure how well it
would work, but it could be interesting.
- Steve
This archive was generated by hypermail 2b28 : Wed Feb 20 2002 - 17:35:02 EET