I'm trying to train a discriminator on L-system outputs. They have the form of a string like "F[/F]FF" with a limited vocabulary of symbols (9 symbols F0[]tTpP/)
I have a bunch of output / bool couple as data (the l system output string, and weither or not this string is "correct").
I'm wondering how many neurons my Input layer need to have ? each string is of variable length, so should the input layer scale as well ?
If you have any clue it would be greatly appreciated.
Thank you
