Evaluation doubt

Discussion of chess software programming and technical issues.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
User avatar
Fabio Gobbato
Posts: 125
Joined: Fri Apr 11, 2014 8:45 am
Contact:

Evaluation doubt

Post by Fabio Gobbato » Sat Oct 29, 2016 5:04 pm

In the evaluation function is it better to have a few weights but well tuned or more weights not well tuned?
What does your experience say?

ZirconiumX
Posts: 1327
Joined: Sun Jul 17, 2011 9:14 am

Re: Evaluation doubt

Post by ZirconiumX » Sat Oct 29, 2016 5:30 pm

Fewer, but better. Fewer weights helps with orthogonality, and the tuning may well make up for the missing terms. Plus, the terms will harmonise themselves with your search, increasing search speed.
Some believe in the almighty dollar.

I believe in the almighty printf statement.

BeyondCritics
Posts: 346
Joined: Sat May 05, 2012 12:48 pm
Location: Bergheim

Re: Evaluation doubt

Post by BeyondCritics » Sun Oct 30, 2016 6:24 am

Read chapter 5 of this book
http://www.deeplearningbook.org/
anf you can dicuss the question yourself. Or read any introduction into "Machine Learning".

Ferdy
Posts: 4111
Joined: Sun Aug 10, 2008 1:15 pm
Location: Philippines

Re: Evaluation doubt

Post by Ferdy » Mon Oct 31, 2016 12:52 pm

BeyondCritics wrote:Read chapter 5 of this book
http://www.deeplearningbook.org/
anf you can dicuss the question yourself. Or read any introduction into "Machine Learning".
Good info, thanks.

User avatar
Fabio Gobbato
Posts: 125
Joined: Fri Apr 11, 2014 8:45 am
Contact:

Re: Evaluation doubt

Post by Fabio Gobbato » Mon Oct 31, 2016 7:20 pm

Of course less weights are easier to tune, but even if they are well tuned they can't express all the knowledge of a more complex evaluation.
So the answer is not so easy.

Post Reply