[freeciv-ai] Re: learning from experience

[Cameron Morland <Cameron@xxxxxxxxxx>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

If you want to implement reinforcement learning, you will need a
better measure of success than simply winning or losing the game. The
learning would be excessively slow otherwise. It would be cheating,
but you could use the game score, even though it isn't the perfect
choice.

Although I don't understand it very well yet, I suspect Adaptive
Resonance Theory could work better, since it can implement fast
learning.

On Fri, Jun 06, 2003 at 08:50:08PM -0400, Alexander L. Vasserman wrote:
> Hi guys,
> I was wondering if anyone has excercised/implemented the idea of AI
> learning from experience. For eample, for the science selection could

- -- 
+-----------------------------------------------------------------
| PGP http://www.eng.uwaterloo.ca/student/cjmorlan/public-key.pgp
| Cameron Morland             ----             Cameron@xxxxxxxxxx
|
| Never argue with a fool, bystanders
| may not be able to tell the difference.
+-----------------------------------------------------------------
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)

iD8DBQE+4eYndgxYKrDrUdgRAkFGAJwLvlaRKUj6bIW9QS/6qcAyV7zNhwCghzns
9VneAthbxY4tialEENAtZNE=
=1HwW
-----END PGP SIGNATURE-----