Every ordering CSP is approximation resistant in the streaming setting
Speaker: NoahTitle: Every ordering CSP is approximation resistant in the streaming setting
Date: 22 Apr 2021 17:00-18:00 EST
Location: Zoom
Food: Self-prepared
Abstract: Maximum ordering constraint satisfaction problems (Max-OCSPs) are optimization problems in which the solution space is the set of all permutations and constraints specify allowed orderings of subsets of the variables; examples include the maximum acyclic subgraph (MAS) and betweenness problems. In 2011, Guruswami, Håstad, Manokaran, Raghavendra, and Charikar showed that, assuming the unique-games conjecture, all Max-OCSPs are classically approximation-resistant. We prove the same result (i.e., the approximation-resistance of all Max-OCSPs) unconditionally in the streaming setting, ultimately invoking streaming hardness results for (non-ordering) CSPs due to recent results of Chou, Golovnev, Sudan, and Velusamy. Previously, no streaming lower bounds were known for any Max-OCSP aside from MAS; for MAS, (¾+\epsilon)-inapproximability for \epsilon>0 was proven by Guruswami and Tao in 2019 using streaming reductions from unique games (while we prove (½+\epsilon)-inapproximability, which is optimal). In this talk, we focus specifically on the new streaming hardness result for MAS and how we use streaming hardness results for CSPs due to Chou, Golovnev, Sudan, and Velusamy.
This is a joint work with Santhoshini Velusamy and Madhu Sudan.