Search found 169 matches

by Rémi Coulom
Mon Nov 28, 2022 4:20 am
Forum: Go AI
Topic: Symplect Go won UEC Cup 14th
Replies: 1
Views: 42

Re: Symplect Go won UEC Cup 14th

Thanks a lot for posting the result and managing the server. I am looking forward to my next participation.
by Rémi Coulom
Sat Nov 19, 2022 12:00 pm
Forum: MCTS, RL, and neural networks
Topic: Experiments with Hex in OpenSpiel and AlphaZero
Replies: 0
Views: 304

Experiments with Hex in OpenSpiel and AlphaZero

https://era.library.ualberta.ca/items/72772250-6770-4d5a-ac64-69ee2e0ccaa1 OpenSpiel is an open-source software system for implementing high-performance software players for many different computer games. Hex is a two-player game of perfect information used in a variety of computer games research pr...
by Rémi Coulom
Thu Nov 10, 2022 5:43 am
Forum: General Discussion
Topic: How to learn programming Game Enemy AI in C++
Replies: 1
Views: 1918

Re: How to learn programming Game Enemy AI in C++

Hi, welcome to the forum! I do not know much about books or tutorial about learning what you want to learn, but we may be able to help you if you ask more precise questions to this forum. My first advice would to try to come up with a solution by yourself. Did you try anything already. What is block...
by Rémi Coulom
Sat Nov 05, 2022 1:03 pm
Forum: Go AI
Topic: beating KataGo by Adversarial attack
Replies: 2
Views: 255

Re: beating KataGo by Adversarial attack

Thanks, these are interesting results. I have always felt that it is a fundamental weakness of the AlphaZero approach that the training set is made of self-play games: it does not learn to handle positions that do not occur between very strong players. That may be why Leela Zero had a lot of trouble...
by Rémi Coulom
Wed Nov 02, 2022 10:58 pm
Forum: ICGA
Topic: Free registration to the Computers and Games Conference (Nov. 22-24)
Replies: 0
Views: 3575

Free registration to the Computers and Games Conference (Nov. 22-24)

Our biennial Computers and Games conference is November 22-24. Registration is free (information is here ). The conference is online and features four keynote talks and 15 research papers. The keynote speakers are: Murray Campbell (IBM TJ Watson) Matt Ginsberg (Google X) Olivier Teytaud (Meta AI Re...
by Rémi Coulom
Wed Nov 02, 2022 10:53 pm
Forum: Go AI
Topic: The 14th UEC Cup will take place in November
Replies: 4
Views: 504

Re: The 14th UEC Cup will take place in November

Great, thanks. The UEC Cup might be an opportunity to advertise this forum to participants. What do you think? I have a bit lost hope that this forum could become really active, but we have made only little publicity, and it might be worth trying a bit more. I have thought about running some kind of...
by Rémi Coulom
Sat Oct 08, 2022 11:48 am
Forum: MCTS, RL, and neural networks
Topic: Tesis ideas about chess programming
Replies: 1
Views: 338

Re: Tesis ideas about chess programming

Hi, Welcome to the forum. If your academic advisor cannot provide you with a good topic, then I think the best approach is to start by studying something you like, and as you grow familiar with the domain, you'll be able to come up with original ideas of things to do by yourself. Writing your own en...
by Rémi Coulom
Tue Oct 04, 2022 8:38 pm
Forum: MCTS, RL, and neural networks
Topic: Scaling Laws for a Multi-Agent Reinforcement Learning Model
Replies: 0
Views: 168

Scaling Laws for a Multi-Agent Reinforcement Learning Model

https://arxiv.org/abs/2210.00849 The recent observation of neural power-law scaling relations has made a significant impact in the field of deep learning. A substantial amount of attention has been dedicated as a consequence to the description of scaling laws, although mostly for supervised learning...
by Rémi Coulom
Sat Jul 30, 2022 11:10 pm
Forum: MCTS, RL, and neural networks
Topic: Paper: U-Net for Gomoku (2021)
Replies: 0
Views: 525

Paper: U-Net for Gomoku (2021)

GomokuNet: A Novel UNet-style Network for Gomoku Zero Learning via Exploiting Positional Information and Multiscale Features Since the tremendous success of the AlphaGo family (AlphaGo, AlphaGo Zero, and Alpha Zero), zero learning has become the baseline method for many board games, such as Gomoku ...