Re: DungeonMaker 1.0 released

[Henningsen <alg@glinx.com>]

Mark Collins wrote:

>Not really a chunk of code....

Indeed not, but it's not a genetic algorithm either. Granted, you use
parameters you call genes, but the essence of the genetic algorithm is that
it stores parents with high fitness value, and generates new gene-sets from
two parents using crossover and mutation. You do none of this.

Any implementation of the genetic algorithm will have to store a sizeable
genepool to file, it will have to completely rewrite that file each time a
new potential parent is added to it, and it will have to allow all possible
integer ranges for its individual genes. Every time a new set of genes is
requested from the engine, it must be generated from two parents read from
file, with crossover and mutation applied to them. In a general
implementation I would also want the concept of chromosomes, and genes with
floating point values, though that isn't strictly required. It is also
neccessary to allow for the tuning of parameters such as mutation rate. In
the end, this will be a solid chunk of code indeed, at least compared to the
small benefits it would provide for automatic difficulty balancing (which
was the context I used the term in).

Frankly, Mark, if you don't even know what a genetic algorithm is, I think
you should reconsider writing a book about that subject. Or at least study
the existing literature some before you do.

Peter Henningsen


---------------------------------------------------------------------
To unsubscribe, e-mail: linuxgames-unsubscribe@sunsite.dk
For additional commands, e-mail: linuxgames-help@sunsite.dk