English

Algorithmic Thresholds for Refuting Random Polynomial Systems

Computational Complexity 2021-10-19 v1 Data Structures and Algorithms

Abstract

Consider a system of mm polynomial equations {pi(x)=bi}im\{p_i(x) = b_i\}_{i \leq m} of degree D2D\geq 2 in nn-dimensional variable xRnx \in \mathbb{R}^n such that each coefficient of every pip_i and bib_is are chosen at random and independently from some continuous distribution. We study the basic question of determining the smallest mm -- the algorithmic threshold -- for which efficient algorithms can find refutations (i.e. certificates of unsatisfiability) for such systems. This setting generalizes problems such as refuting random SAT instances, low-rank matrix sensing and certifying pseudo-randomness of Goldreich's candidate generators and generalizations. We show that for every dNd \in \mathbb{N}, the (n+m)O(d)(n+m)^{O(d)}-time canonical sum-of-squares (SoS) relaxation refutes such a system with high probability whenever mO(n)(nd)D1m \geq O(n) \cdot (\frac{n}{d})^{D-1}. We prove a lower bound in the restricted low-degree polynomial model of computation which suggests that this trade-off between SoS degree and the number of equations is nearly tight for all dd. We also confirm the predictions of this lower bound in a limited setting by showing a lower bound on the canonical degree-44 sum-of-squares relaxation for refuting random quadratic polynomials. Together, our results provide evidence for an algorithmic threshold for the problem at mO~(n)n(1δ)(D1)m \gtrsim \widetilde{O}(n) \cdot n^{(1-\delta)(D-1)} for 2nδ2^{n^{\delta}}-time algorithms for all δ\delta.

Keywords

Cite

@article{arxiv.2110.08677,
  title  = {Algorithmic Thresholds for Refuting Random Polynomial Systems},
  author = {Jun-Ting Hsieh and Pravesh K. Kothari},
  journal= {arXiv preprint arXiv:2110.08677},
  year   = {2021}
}