Re: [Computer-go] CGOS source on github

DeepMind has published a number of papers on how to stabilize RL strategies in a landscape of nontransitive cycles. See https://papers.nips.cc/paper/2018/file/cdf1035c34ec380218a8cc9a43d438f9-Paper.pdf I haven't fully digested the paper, but what I'm getting from it is that if you want your

Re: [Computer-go] Crazy Stone is playing on CGOS 9x9

Agreed on the book interface! If you click through to the end of a variation, you can see that the evaluation is a minimax'd evaluation, instead of an average over the subtree nodes, even if the overall tree was generated via a UCT algorithm. The values seem a bit more fragile to me as a

Re: [Computer-go] mit-develops-algorithm-to-accelerate-neural-networks-by-200x

this doesn't actually speed up the neural networks that much; it's a technique to more quickly brute-force the search space of possible neural networks for ones that execute faster while maintaining similar accuracy. Typical hype article. Anyway, the effort spent looking for bizarre architectures

Re: [Computer-go] Neural networks + MCTS applied to chemical syntheses

In particular, they had no way to train a value net, so it was back to AlphaGo v1 style of training just a policy net and reusing it as the rollout policy. On Fri, Apr 6, 2018 at 6:31 AM Fidel Santiago wrote: > Hello, > > Apparently the lessons of Alphago (and many

[Computer-go] Hosting mid-level bots on CGOS

Hi everyone, I remember that when I first got started out and put a fledgling MuGo on CGOS, I was discouraged at the vast gulf between GnuGo and the high-dan bots. Now that CGOS is flooded with all sorts of NN-based bots that have very similar playstyles and weaknesses, the Minigo team has been

[Computer-go] Minigo's 9x9 training data available for download

Hi everyone, The artifacts of our latest 9x9 run are publicly available for download at gs://minigo-pub/v3-9x9 Under this directory you can find data/ Tensorflow tf.Record files corresponding to raw training data models/ The saved model files for each generation sgf/ Zip archives of all games in

[Computer-go] MiniGo open sourced

Hi everybody, I'm happy to announce MiniGo is now open source. https://github.com/tensorflow/minigo We're not aiming for a top-level Go AI; we're merely aiming for a correct, very readable implementation of the AlphaGoZero algorithm and demonstration of Google Cloud / Kubernetes / TensorFlow.

Re: [Computer-go] action-value Q for unexpanded nodes

It should default to the Q of the parent node. Otherwise, let's say that the root node is a losing position. Upon choosing a followup move, the Q will be updated to a very negative value, and that node won't get explored again - at least until all 362 top-level children have been explored and

[Computer-go] purpose of zero-filled feature planes in CNN

I've been wondering about something I've seen in a few papers (AlphaGo's paper, Cazenave's resnet policy architecture), which is the presence of an input plane filled with 0s. The input features also typically include a plane of 1s, which makes sense to me - zero-padding before a convolution

[Computer-go] World AI Go Open

Hello everyone, Is anybody else planning to go to this event? https://www.reddit.com/r/baduk/comments/6ge1ul/the_1st_world_ai_go_open/ I'm a first timer to in-person computer go tournaments and am unsure how I should deal with logistics. I don't want to pack up my desktop in a suitcase so I've

Re: [Computer-go] Computer-go Digest, Vol 83, Issue 7

As I understand it, RL boosts the performance of the policy network, not because the winner's moves are particularly better than the loser's moves, but because it shores up specifically the weaknesses of the SL-trained network. In other words, a network trained purely with SL will have certain

Re: [Computer-go] Converging to 57%

I've been working on my own AlphaGo replication (code on github https://github.com/brilee/MuGo), and I've found it reasonably easy to hit 45% prediction rate with basic features (stone locations, liberty counts, and turns since last move), and a relatively small network (6 intermediate layers, 32