arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1602.02743cs.LOcs.AIcs.CCcs.FL
收藏

IMP游戏:超越$Σ^0_1$的可学习性、可逼近性和对抗性学习

The IMP game: Learnability, approximability and adversarial learning beyond $Σ^0_1$

Monash University(莫纳什大学)

Michael Brand, David L. Dowe

摘要:

我们介绍了一个问题设置,我们称之为迭代匹配便士(IMP)游戏,并表明它是研究三个问题的强大框架:对抗性学习,传统(即,非对抗性)可学习性和可逼近性。利用它,我们能够导出下列定理。(1)通过示例可以学习所有的$Σ^0_1 \cup Π^0_1$以及一些超集;(2)在对抗学习(我们将其描述为追求-逃避游戏)中,追求者具有获胜策略(换句话说,$Σ^0_1$可以通过对抗学习,但$Π^0_1$不能);(3)$Π^0_1$中的某些语言不能用$Σ^0_1$中的任何语言来近似。对于任意$i$,我们也给出了$Σ^0_i$和$Π^0_i$的相应结果。

英文摘要:

We introduce a problem set-up we call the Iterated Matching Pennies (IMP) game and show that it is a powerful framework for the study of three problems: adversarial learnability, conventional (i.e., non-adversarial) learnability and approximability. Using it, we are able to derive the following theorems. (1) It is possible to learn by example all of $Σ^0_1 \cup Π^0_1$ as well as some supersets; (2) in adversarial learning (which we describe as a pursuit-evasion game), the pursuer has a winning strategy (in other words, $Σ^0_1$ can be learned adversarially, but $Π^0_1$ not); (3) some languages in $Π^0_1$ cannot be approximated by any language in $Σ^0_1$. We show corresponding results also for $Σ^0_i$ and $Π^0_i$ for arbitrary $i$.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑