By Junichi Hashimoto, Akihiro Kishimoto, Kazuki Yoshizoe (auth.), H. Jaap van den Herik, Aske Plaat (eds.)
This e-book constitutes the completely refereed post-conference complaints of the thirteenth Advances in laptop video games convention, ACG 2011, held in Tilburg, The Netherlands, in November 2011. The 29 revised complete papers awarded have been rigorously reviewed and chosen from a variety of submissions. The papers conceal quite a lot of subject matters corresponding to Monte-Carlo tree seek and its enhancement, temporal distinction studying, optimization, fixing and looking, research of a online game attribute, new ways, and critical games.
Read or Download Advances in Computer Games: 13th International Conference, ACG 2011, Tilburg, The Netherlands, November 20-22, 2011, Revised Selected Papers PDF
Similar international books
This ebook constitutes the completely refereed post-proceedings of the fifth foreign Workshop on severe details Infrastructure defense, CRITIS 2010, held in Athens, Greece in September 2010. The 12 revised complete papers and poster papers offered went via rounds of reviewing and development and have been chosen from 30 submissions.
A contemporary accomplished research of tension and human future health via the Institute of Medicine/National Academy of Sciences concludes that folks who event any of quite a lot of tense occasions or events are at elevated probability of constructing a actual or psychological affliction, together with middle ailment.
- Quantum Interaction: 6th International Symposium, QI 2012, Paris, France, June 27-29, 2012, Revised Selected Papers
- International Handbook of Mathematics Education: Part 1
- Acidic Precipitation: Proceedings of the International Symposium on Acidic Precipitation Muskoka, Ontario, September 15–20, 1985
- Scale Space and Variational Methods in Computer Vision: 4th International Conference, SSVM 2013, Schloss Seggau, Leibnitz, Austria, June 2-6, 2013. Proceedings
- Human-Computer Interaction – INTERACT 2013: 14th IFIP TC 13 International Conference, Cape Town, South Africa, September 2-6, 2013, Proceedings, Part I
Additional info for Advances in Computer Games: 13th International Conference, ACG 2011, Tilburg, The Netherlands, November 20-22, 2011, Revised Selected Papers
We terminate search early if the most simulated node cannot change even if it did not receive any more simulations for the rest of the tm time. We used a very fast “low-end” scenario with 300 seconds per game. PACHI: State of the Art Open Source Go Program 31 allocation (adjusting time non-linearly, for a peak, not a plateau). Our middle game time allocation algorithm is not a source of signiﬁcant performance beneﬁts and we expect that our and Erica’s algorithm could be reconciled. 2 Dynamic Komi The MCTS algorithm evaluates the possible moves most accurately when the winning rate is near 50%.
Self-play scalability (ﬁxed sims) Fig. , the increase in playing time needed to achieve identical strength play [6,20]. The tests are done using the “high-end” scenario, but Pachi uses a variable number of cores. 5. 2 Cluster Parallelization The MCTS cluster parallelization is still far from being clearly solved. Pachi features elaborate support for distributed computations with information exchange between the nodes, but it still scales much slower when multiplying the number of nodes rather than processors with a low-latency shared tree.
Algorithm 1. NodeValue Require: sims, simsAMAF are numbers of simulations pertaining the node. Require: wins, winsAMAF are numbers of won simulations. NormalTerm ← wins sims RAVE AMAF RAVETerm ← wins = wins simsRAVE simsAMAF simsRAVE β ← simsRAVE +sims+sims RAVE ·sims/3000 NodeValue ← (1 − β) · NormalTerm + β · RAVETerm The MCTS variants diﬀer in the choice of the next node during the descent. Pachi uses the RAVE algorithm  that takes into account not only per-child winrate statistics for the move being played next during the descent, but also (as a separate value) anytime later5 during the simulation (the so-called AMAF statistics).