-
https://research.fb.com/facebook-open-sources-elf-opengo/
ELF OpenGo
ELF OpenGo is a reimplementation of AlphaGoZero / AlphaZero. It was trained on 2,000 GPUs over a two week period, and has achie…
-
I made some analysis earlier that maybe lowering visits would be useful but it wasn't very rigorous (https://github.com/gcp/leela-zero/issues/1030#issuecomment-374328038, https://github.com/gcp/leela-…
-
Please find attach an Excel file simulating the impact of gating vs no gating, to try to understand the most efficient approach.
Last version of the Excel (18/6/2018):
[Gating v4.xlsx.gz](https://g…
-
Hello!
Since it required huge amount of computing resources to teach network - is it possible to create some distributed system, where everyone who is willing can join and contribute their machine r…
-
**Synthesis (after two updates on 27/11/2018):**
**ALL match games** tested since yesterday (~80 games from 3 tested nets against LZ191) **have exactly the same opening for the first 30 moves!** For…
-
As you know, there are some different approaches to utilize value network's evaluation.
For example, An open sourced computer go program RN, just averages value network evaluations and uses it, lik…
-
Since a lot of people are working on tuning FPU at the moment and some people are exploring tweaks to the search algorithm I wanted to share a few areas of research I was looking over this evening, in…
-
is the aim to produce much stronger with weight bigger size or not so much stronger and keep the size increases small ?
and also care about the time of self play to get promotion?
some possible size…
-
3582422213.03M+128k40x256 VS 85a93684 | 243 : 159 (60.45%)
What does it mean? Several new promoted nets over several weeks, yielded 60% progress?
Is it rock/paper/scissors issue?
-
*Upon suggestion of @roy7 , I post here a discussion held on discord today*
@Friday9i Regarding the question why opening move policy tends to become almost 100% on one move (up to the 8 symetries),…