Computer Science > Computational Complexity
[Submitted on 14 Apr 2023]
Title:Solving Unique Games over Globally Hypercontractive Graphs
View PDFAbstract:We study the complexity of affine Unique-Games (UG) over globally hypercontractive graphs, which are graphs that are not small set expanders but admit a useful and succinct characterization of all small sets that violate the small-set expansion property. This class of graphs includes the Johnson and Grassmann graphs, which have played a pivotal role in recent PCP constructions for UG, and their generalizations via high-dimensional expanders.
Our algorithm shows how to round "low-entropy" solutions to sum-of-squares (SoS) semidefinite programs, broadly extending the algorithmic framework of [BBKSS'21]. We give a new rounding scheme for SoS, which eliminates global correlations in a given pseudodistribution so that it retains various good properties even after conditioning. Getting structural control over a pseudodistribution after conditioning is a fundamental challenge in many SoS based algorithms. Due to these challenges, [BBKSS] were not able to establish strong algorithms for globally hypercontractive graphs, and could only do so for certifiable small-set expanders. Our results improve upon the results of [BBKSS] in various aspects: we are able to deal with instances with arbitrarily small (but constant) completeness, and most importantly, their algorithm gets a soundness guarantee that degrades with other parameters of the graph (which in all PCP constructions grow with the alphabet size), whereas our doesn't.
Our result suggests that UG is easy on globally hypercontractive graphs, and therefore highlights the importance of graphs that lack such a characterization in the context of PCP reductions for UG.
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
Connected Papers (What is Connected Papers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.