Neuronet plus conventional approach combined?

Discussion of chess software programming and technical issues.

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
Ras
Posts: 1140
Joined: Tue Aug 30, 2016 6:19 pm
Contact:

Neuronet plus conventional approach combined?

Post by Ras » Fri Sep 02, 2016 8:03 pm

A question that I've been thinking about for a while:

Neuronets alone suffer from tactical problems because these networks are designed to answer minimum changes in the input with minimum changes in the output. Chess is a game where displacing one piece by one square can make a decisive difference, so that doesn't always blend well.

But what about an engine that does a conventional tree search with a minimum eval just to filter all tactically sound moves and feeds them to a neuronet for the final selection?

Are there engines that implement this approach? Or is the whole idea nonsense?

matthewlai
Posts: 791
Joined: Sun Aug 03, 2014 2:48 am
Location: London, UK
Contact:

Re: Neuronet plus conventional approach combined?

Post by matthewlai » Fri Sep 02, 2016 8:18 pm

Ras wrote:A question that I've been thinking about for a while:

Neuronets alone suffer from tactical problems because these networks are designed to answer minimum changes in the input with minimum changes in the output. Chess is a game where displacing one piece by one square can make a decisive difference, so that doesn't always blend well.

But what about an engine that does a conventional tree search with a minimum eval just to filter all tactically sound moves and feeds them to a neuronet for the final selection?

Are there engines that implement this approach? Or is the whole idea nonsense?
Giraffe uses a conventional search with a neural net evaluation function. Using a fast search to sort out tactics is another approach that I have spent quite some time thinking about, but haven't had a chance to implement.
Disclosure: I work for DeepMind on the AlphaZero project, but everything I say here is personal opinion and does not reflect the views of DeepMind / Alphabet.

Ras
Posts: 1140
Joined: Tue Aug 30, 2016 6:19 pm
Contact:

Re: Neuronet plus conventional approach combined?

Post by Ras » Fri Sep 02, 2016 10:01 pm

Thanks, that's an answer from the chess-neuro-expert No. 1.

I mean, computers are better than humans with their current tactics approach, but strategically, human masters are still better. Only that humans fail tactically before their strategies might come to fruitition.

What is your take, are neuronets possibly suited better to the diffuse strategic long-term aspects of the game? Could they implement this in case they don't have to worry about tactical shots?

thomasahle
Posts: 71
Joined: Thu Feb 27, 2014 7:19 pm

Re: Neuronet plus conventional approach combined?

Post by thomasahle » Fri Sep 02, 2016 10:34 pm

Spawkfish is an engine that tries to predict moves only with neural nets, and does reasonably well at that. If it could be made fast enough, using it for move ordering and heavy LMR would probably make a quite strong engine.

matthewlai
Posts: 791
Joined: Sun Aug 03, 2014 2:48 am
Location: London, UK
Contact:

Re: Neuronet plus conventional approach combined?

Post by matthewlai » Fri Sep 02, 2016 11:07 pm

Ras wrote:Thanks, that's an answer from the chess-neuro-expert No. 1.

I mean, computers are better than humans with their current tactics approach, but strategically, human masters are still better. Only that humans fail tactically before their strategies might come to fruitition.

What is your take, are neuronets possibly suited better to the diffuse strategic long-term aspects of the game? Could they implement this in case they don't have to worry about tactical shots?
It's an interesting idea. Neural nets are certainly better at positional stuff than what we have at the moment, but they are a bit too slow to compete tactically.

There are some practical issues with using 2 evaluation functions in the same search tree, so that will still require some thinking. I have some ideas on how to do it, but nothing I am really confident in yet.
Disclosure: I work for DeepMind on the AlphaZero project, but everything I say here is personal opinion and does not reflect the views of DeepMind / Alphabet.

matthewlai
Posts: 791
Joined: Sun Aug 03, 2014 2:48 am
Location: London, UK
Contact:

Re: Neuronet plus conventional approach combined?

Post by matthewlai » Fri Sep 02, 2016 11:12 pm

thomasahle wrote:Spawkfish is an engine that tries to predict moves only with neural nets, and does reasonably well at that. If it could be made fast enough, using it for move ordering and heavy LMR would probably make a quite strong engine.
Giraffe already does that. It's called the "move evaluator" net in Giraffe. It didn't help as much as I had hoped, but it did help a bit.
Disclosure: I work for DeepMind on the AlphaZero project, but everything I say here is personal opinion and does not reflect the views of DeepMind / Alphabet.

thomasahle
Posts: 71
Joined: Thu Feb 27, 2014 7:19 pm

Re: Neuronet plus conventional approach combined?

Post by thomasahle » Sun Sep 04, 2016 10:26 pm

matthewlai wrote:
thomasahle wrote:Spawkfish is an engine that tries to predict moves only with neural nets, and does reasonably well at that. If it could be made fast enough, using it for move ordering and heavy LMR would probably make a quite strong engine.
Giraffe already does that. It's called the "move evaluator" net in Giraffe. It didn't help as much as I had hoped, but it did help a bit.
Cool! How strong would you estimate the move evaluator is on its own, with no search?

matthewlai
Posts: 791
Joined: Sun Aug 03, 2014 2:48 am
Location: London, UK
Contact:

Re: Neuronet plus conventional approach combined?

Post by matthewlai » Sun Sep 04, 2016 10:49 pm

thomasahle wrote:
matthewlai wrote:
thomasahle wrote:Spawkfish is an engine that tries to predict moves only with neural nets, and does reasonably well at that. If it could be made fast enough, using it for move ordering and heavy LMR would probably make a quite strong engine.
Giraffe already does that. It's called the "move evaluator" net in Giraffe. It didn't help as much as I had hoped, but it did help a bit.
Cool! How strong would you estimate the move evaluator is on its own, with no search?
Depends on what features you give it. If you give it SEE scores maybe 2000 level. If you don't, it will probably make many tactical mistakes.
Disclosure: I work for DeepMind on the AlphaZero project, but everything I say here is personal opinion and does not reflect the views of DeepMind / Alphabet.

thomasahle
Posts: 71
Joined: Thu Feb 27, 2014 7:19 pm

Re: Neuronet plus conventional approach combined?

Post by thomasahle » Mon Sep 05, 2016 1:30 pm

matthewlai wrote:Depends on what features you give it. If you give it SEE scores maybe 2000 level. If you don't, it will probably make many tactical mistakes.
Sounds like something that would be fun to play against. If I run giraffe and tell it to search depth 0, will that simulate the policy network reasonably well?

matthewlai
Posts: 791
Joined: Sun Aug 03, 2014 2:48 am
Location: London, UK
Contact:

Re: Neuronet plus conventional approach combined?

Post by matthewlai » Mon Sep 05, 2016 1:51 pm

thomasahle wrote:
matthewlai wrote:Depends on what features you give it. If you give it SEE scores maybe 2000 level. If you don't, it will probably make many tactical mistakes.
Sounds like something that would be fun to play against. If I run giraffe and tell it to search depth 0, will that simulate the policy network reasonably well?
That wouldn't work unfortunately. It would still do a QSearch. The move evaluator network is also only trained on quiet moves.
Disclosure: I work for DeepMind on the AlphaZero project, but everything I say here is personal opinion and does not reflect the views of DeepMind / Alphabet.

Post Reply