Tight lower bounds for the complexity of multicoloring

Abstract: In the multicoloring problem, also known as ($a$:$b$)-coloring or $b$-fold coloring, we are given a graph G and a set of $a$ colors, and the task is to assign a subset of $b$ colors to each vertex of G so that adjacent vertices receive disjoint color subsets. This natural generalization of the classic coloring problem (the $b=1$ case) is equivalent to finding a homomorphism to the Kneser graph $KG_{a,b}$, and gives relaxations approaching the fractional chromatic number. We study the complexity of determining whether a graph has an ($a$:$b$)-coloring. Our main result is that this problem does not admit an algorithm with running time $f(b)\cdot 2^{o(\log b)\cdot n}$, for any computable $f(b)$, unless the Exponential Time Hypothesis (ETH) fails. A $(b+1)^n\cdot \text{poly}(n)$-time algorithm due to Nederlof [2008] shows that this is tight. A direct corollary of our result is that the graph homomorphism problem does not admit a $2<sup>{O(n+h)}$ algorithm unless ETH fails, even if the target graph is required to be a Kneser graph. This refines the understanding given by the recent lower bound of Cygan et al. [SODA 2016]. The crucial ingredient in our hardness reduction is the usage of detecting matrices of Lindstr\"om [Canad. Math. Bull., 1965], which is a combinatorial tool that, to the best of our knowledge, has not yet been used for proving complexity lower bounds. As a side result, we prove that the running time of the algorithms of Abasi et al. [MFCS 2014] and of Gabizon et al. [ESA 2015] for the r-monomial detection problem are optimal under ETH.