Search found 31 matches

by grahamj
Wed Mar 20, 2019 2:58 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Minimax, noisy evaluations, PUCT
Replies: 1
Views: 568

Minimax, noisy evaluations, PUCT

Minimax is known to have a bias which does not diminish with depth. The paper Bias and pathology in minimax search , 2005, A. Sadikov, I. Bratko, I. Kononenko uses KRK endings to investigate this. Here's a simplified situation which I was able to analyse mathematically. Suppose the branching factor ...
by grahamj
Sun Mar 03, 2019 4:04 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Training the trainer: how is it done for Stockfish?
Replies: 22
Views: 2937

Re: Training the trainer: how is it done for Stockfish?

On why AlphaZero (and hence LC0) opted for a Monte Carlo search... There are no strong chess engines which use alpha-beta pruning alone. All the strong engines use other techniques to do a huge amount of pruning. These other techniques are mostly chess-specific, so no good for DeepMind's purposes. A...
by grahamj
Thu Feb 21, 2019 7:41 am
Forum: Computer Chess Club: General Topics
Topic: superhuman engines on common hardware?
Replies: 16
Views: 2679

Re: superhuman engines on common hardware?

mwyoung wrote:
Thu Feb 21, 2019 2:17 am
Maybe I don't understand the question. But it seems to me we are already running superhuman NN engines on common hardware. All I had to do was go to Micro Center.
Maybe you didn't notice the date of the original post.
by grahamj
Tue Feb 19, 2019 8:41 am
Forum: Computer Chess Club: General Topics
Topic: Lc0 ... the GOOD and the BAD
Replies: 54
Views: 7105

Re: Lc0 ... the GOOD and the BAD

I think Monte Carlo search should actually be much stronger in tactical suites than AB engines as it can cover many more positions quicker... How do you imagine that might work? What algorithm do you have in mind? Oh, I would have no clue about how to do that. But I think that there is very big pot...
by grahamj
Mon Feb 18, 2019 5:48 pm
Forum: Computer Chess Club: General Topics
Topic: Lc0 ... the GOOD and the BAD
Replies: 54
Views: 7105

Re: Lc0 ... the GOOD and the BAD

M ANSARI wrote:
Mon Feb 18, 2019 2:09 pm
I think Monte Carlo search should actually be much stronger in tactical suites than AB engines as it can cover many more positions quicker...
How do you imagine that might work? What algorithm do you have in mind?
by grahamj
Mon Feb 11, 2019 9:36 pm
Forum: Computer Chess Club: General Topics
Topic: CUDA benchmark and Elo of Leela
Replies: 3
Views: 691

Re: CUDA benchmark and Elo of Leela

There are some scaling test results here, from last August.
https://docs.google.com/spreadsheets/d/ ... 1017575449
by grahamj
Sat Feb 09, 2019 2:44 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: WIP, Eta - GPGPU ANN based engine, RFC
Replies: 10
Views: 1836

Re: WIP, Eta - GPGPU ANN based engine, RFC

With 'zero cost' i mean adding a qsearch to LC0 without any additional computation costs, no nps drop or similar, as thought experiment. -- Srdja I don't know enough to comment on your search algorithm. I can suggest a real experiment which might help you evaluate your q-search idea. In http://talk...
by grahamj
Sat Feb 09, 2019 8:54 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: WIP, Eta - GPGPU ANN based engine, RFC
Replies: 10
Views: 1836

Re: WIP, Eta - GPGPU ANN based engine, RFC

I hope the Q-Search at leafnodes will do some magic. If the NN is only called to evaluate quiescent positions, it should mostly, perhaps entirely, be trained on quiescent positions. It ciould be tricky to collect suitable training data, especially If your definition of quiescent is complicated. Peo...
by grahamj
Wed Feb 06, 2019 6:43 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: WIP, Eta - GPGPU ANN based engine, RFC
Replies: 10
Views: 1836

Re: WIP, Eta - GPGPU ANN based engine, RFC

I hope the Q-Search at leafnodes will do some magic. If the NN is only called to evaluate quiescent positions, it should mostly, perhaps entirely, be trained on quiescent positions. It ciould be tricky to collect suitable training data, especially If your definition of quiescent is complicated. Peo...
by grahamj
Wed Feb 06, 2019 4:29 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: chess programming - predictions for the next 5 years
Replies: 15
Views: 2511

Re: chess programming - predictions for the next 5 years

I don't think anyone will be able to "unblackbox" neural networks in 5 years. ML researchers would love to be able to do this, but I predict very slow progress in general, and almost none in chess. I shouldn't have said "Interest in neural nets will wane...", what I really meant was "deep convolutio...