推荐dharma1 · 2018-05
“This paper was super interesting - [OTHER_PAPER], as was this - A Generalised Method for Empirical Game Theoretic Analysis”
称这篇论文非常有意思,与另一篇一并推荐
给复杂多智能体交互的经验元博弈分析提供理论界与非对称扩展
论文为复杂多智能体交互的经验博弈论分析给出理论界:元博弈的纳什均衡是真实底层博弈的近似纳什均衡,并研究了得到足够好近似所需的数据样本量。作者还把元博弈分析方法扩展到非对称博弈,并在多个 AlphaGo 版本的进化动态(对称)、Facebook 上人类玩家的 Colonel Blotto 博弈(对称)以及 PSRO 生成的 Leduc 扑克元博弈(非对称)中做了经验演示。
HN 评论者觉得这篇论文非常有意思,并引用其结果来说明至少截至当时,AlphaGo 的各个版本之间倾向于传递性而没有循环。
“This paper was super interesting - [OTHER_PAPER], as was this - A Generalised Method for Empirical Game Theoretic Analysis”
“indicate that AlphaGo versions, at least up until then, tend to be transitive:”