Playing Unique Games on Certified Small-Set Expanders
Speaker: MitaliTitle: Playing Unique Games on Certified Small-Set Expanders
Date: 15 Oct 2020 17:15-18:30 EST (talk starts at 17:30)
Location: Zoom
Food: Self-prepared
Abstract: The unique games problem (UG) and the small-set expansion (SSE) problem are well-studied problems in complexity theory and giving a reduction from UG to SSE is an important open problem. The UG problem is a constraint satisfaction problem on alphabet [k], where each constraint is a 2-variate linear equation mod k, whereas the SSE problem asks that given a graph G, find a small non-expanding set in G if it exists.
We give an algorithm for solving unique games (UG) instances whose constraints correspond to edges of graphs that are small-set expanders and furthermore have a sum-of-squares certificate of expansion. Through this algorithm, we partially relate the complexity of the UG and SSE problems. Our techniques can also be extended to solve UG on graphs that are not small set expanders, but have a low-degree characterization of the non-expanding sets. As corollaries, we obtain the first polynomial-time algorithms for solving UG on the noisy hypercube, the short code and Johnson graphs. The prior best algorithm for such instances was the algorithm of [ABS]’10 which required quasi-polynomial time for the noisy hypercube and nearly-exponential time for the short code and Johnson graphs.
In this talk, I will first introduce the UG and SSE problems, conjectures about them, and then give the main ideas to prove our theorems. I won’t assume any prior knowledge of sum-of-squares algorithms. Based on our paper.