Re: An AI question and a UI question

Paul wrote:
> I've been working on my RLG, Fugue, for the past few years, and while
> it hasn't reached the megalomaniacal ambitions of Armok, it's pretty
> hefty for a first try.  Ok, now that I've covered that, here's the
> problem.  I'm looking at doing this with some neural nets in
> strategically placed locations in the AI so that it won't be brain
> dead.  Unfortunately, I've been trying to deal with the problem of
> evaluating the current situation.

Wow, neural networks are cool but I think they need some pre-work before 
you can use them.

Step 1 : choose your inputs : typically that would be
 - the grid that can be seen/is known by the 'monster'
 - percent health monster
 - perceived health hero
 - health difference monster since previous turn
   ( either healing by monster or getting damage by hero )
 - perceived health difference since previous turn
   ( either healing by hero or getting damage by monsters )
 - monster can do distance attacks even when stuff is in the way( bool )
 - monster can do distance attacks if stuff is not in the way( bool )
 - hero can do distance attacks even when stuff is in the way( bool )
 - hero can do distance attacks if stuff is not in the way( bool )

Step 2:
Then you have to make a borg, make it as powerfull as it will generally
be when they meet the monster.

Step 3:
Define fitness condition :
usually go for kill hero without getting killed,
or just not getting killed.

Step 4:
The Darwin Machine :
Run and monitor your neural nets, do random changes to the weights and
inhibitors and raise your neural networks to clever little beings which
will annoy the players.

Step 0 :
Buy the book Blondie24, it gives insight to neural networks and deals
with neural networks analying a grid of checkers, which ( with some
fantasy ) can be used for analyzing a dungeon floor.


> Thanks in advance.
> 
> -- Paul

T.