Aldous–Lyons 猜想 II:不可判定性
The Aldous--Lyons Conjecture II: Undecidability
AI总结:
本文通过证明定制非局部博弈完美策略存在性的不可判定性,并借助归约技术,给出了 Aldous–Lyons 猜想的否定回答,同时以更精简的方式重新证明了 Connes 嵌入问题的否定结论。
摘要:
本文及其姊妹篇 [BCLV24] 致力于对 Aldous–Lyons 猜想 [AL07, Ald07] 作出否定性解决。在本部分中,我们研究定制非局部博弈(tailored non-local games)。这是非局部博弈的一个子类——非局部博弈是用于建模量子力学中某些实验以及复杂性理论中交互式证明的组合对象。我们的主要结果是:给定一个定制非局部博弈 $G$,区分以下两种情形是不可判定的:$G$ 具有一种特殊类型的完美策略的情形,以及 $G$ 的每个策略都远非完美的情形。利用姊妹篇 [BCLV24] 中引入的归约,这一不可判定性结果蕴含了对 Aldous–Lyons 猜想的否定回答。即,它蕴含了非 sofic 的单模网络(unimodular networks)的存在性。为证明我们的结果,我们使用了 MIP*=RE [JNV+21] 中发展的压缩技术的一个变体。我们的主要技术贡献是将该技术适配到定制非局部博弈这一类别。主要困难在于建立答案归约(answer reduction),这需要对概率可检验证明(probabilistically checkable proofs)构造中的现有技术进行非常细致的适配。作为副产品,我们重新证明了 Connes 嵌入问题 [Con76] 的否定——即存在一个不能嵌入到超有限 $\mathrm{II}_1$-因子的超幂中的 $\mathrm{II}_1$-因子——该结果首先在 [JNV+21] 中被证明,而我们使用了一个可以说更为精简的证明。特别地,我们吸收了文献 [dlS22b, Vid22] 中由 de la Salle 和第三作者给出的近期简化。
英文摘要:
This paper, and its companion [BCLV24], are devoted to a negative resolution of the Aldous--Lyons Conjecture [AL07, Ald07]. In this part we study tailored non-local games. This is a subclass of non-local games -- combinatorial objects which model certain experiments in quantum mechanics, as well as interactive proofs in complexity theory. Our main result is that, given a tailored non-local game $G$, it is undecidable to distinguish between the case where $G$ has a special kind of perfect strategy, and the case where every strategy for $G$ is far from being perfect. Using a reduction introduced in the companion paper [BCLV24], this undecidability result implies a negative answer to the Aldous--Lyons conjecture. Namely, it implies the existence of unimodular networks that are non-sofic. To prove our result, we use a variant of the compression technique developed in MIP*=RE [JNV+21]. Our main technical contribution is to adapt this technique to the class of tailored non-local games. The main difficulty is in establishing answer reduction, which requires a very careful adaptation of existing techniques in the construction of probabilistically checkable proofs. As a byproduct, we are reproving the negation of Connes' embedding problem [Con76] -- i.e., the existence of a $\mathrm{II}_1$-factor which cannot be embedded in an ultrapower of the hyperfinite $\mathrm{II}_1$-factor -- first proved in [JNV+21], using an arguably more streamlined proof. In particular, we incorporate recent simplifications from the literature [dlS22b, Vid22] due to de la Salle and the third author.
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。