[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: computer-go: Fast scoring program?



> then it takes it. otherwise it passes. But of course a random net you
> would expect it to pass alot from the beginning.

Not quite. A random net should have about 50% outputs on the positive and
50% in the negative. Thus lots of moves to choose from. 


> Also if its an empty board
> i fill the C3 position with the enemy on the network, cause 
> an empty board will get all 0 out puts in the net. 

I guess that you don't have any input in your net to detect the edge of the
board. You either need to place that, or introduce some predefined (or
trained) background noise that should help you initiate the game.


> Its the fitness function thats the tough problem here, what's 
> the trend you are having?

My fitness function is the result of a tournament withing the population.
What do you mean by trend? You see, the fitness function is a function of an
individual in a population. The relationship between fitnesses of two
individuals is irrelevant if the individuals are evaluated in different
populations. The most I could do is have the best of each generation play
against best in prior generations and in general the more evolved 'moron'
should win. I call it 'moron' for based on my results, they do play like
morons. Yet I haven't done such test yet.


> I've played with deducting points on passing depending on
> empty spaces (was score at that time, but that was too slow) 
> but i rarely got above 20 moves per game and averages 8
> moves.. way too low on a 9x9 board.

I had absolutely no problem with passing too early. On the contrary, most of
my nets never quit. I had to artificially end games after about 500 moves.
As I said, a random net should propose to you plenty of moves. Maybe your
encoding or interpretation of that encoding is skewed toward passing. You
need to revisit these areas and make sure that doesn't happen.


> Maybe i can use gnugo's scoreing function to my advantage and
> subtract the time it takes to score a game from the final 
> score. On the
> asumption if its  a long time to score then there is a random
> distrabution. And maybe subtract the time from the loser only. Hmm
> actually thats what i'll do have nothing to lose at this point...

As I said, I had problems with some random distributions never finish
scoring in gnugo. I would verify why I get the passing so often and solve it
from that side.
Keep me posted.

Good luck,
Adrian