Re: GA Music fitness: a practical guide


dvunkannon@xxxxxxxxxxx wrote:
Your difficulty has nothing to do with using a non-binary alphabet.
Your fitness function is too coarse, just {good, bad}. Even scoring the
melodies 1 to 10, 1= bad, 10= great, would be an improvement.

BTW, you might try changing your representation. Instead of the alleles
representing notes, you might try letting them represent the intervals
between successive notes. In your current representation, all melodies
are trapped in one octave. Using an interval representation, a 20 note
string could cover several octaves. It would also let you transpose the
melody into different keys easily. For example, in an interval
representation, 1 = -8 intervals, 9 = same note, 17 = +8 intervals. A
smaller alphabet would allow less of a jump from one note to the next.

Thanks really much for your help,but my main question is clear:

Supposing that I like a lot two schemes only ("2 2" "3 6 3 2" or any
two of different length)
how can i make them appear in the fittest chromosome after some
generations with predetermined ratio (70-30,60-40 or other)?

.

Relevant Pages

  • Re: GA Music fitness: a practical guide
    ... Your difficulty has nothing to do with using a non-binary alphabet. ... BTW, you might try changing your representation. ... are trapped in one octave. ...
    (comp.ai.genetic)
  • Re: Gonna or Gonna
    ... BTW, in mysteries written by an Australian (whose name I cannot ... That's a fairly good representation of the pronunciation. ...
    (alt.usage.english)