Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Computational Lower Bounds for Graphon Estimation via Low-degree Polynomials

Yuetian Luo, Chao Gao

TL;DR

This work establishes rigorous computational lower bounds for graphon estimation by applying the low-degree polynomial framework to SBM-based priors and nonparametric graphons. It proves that, for degree-$D$ polynomial estimators, the estimation risk cannot beat the USVT benchmark up to polylog factors in the SBM regime ($\ge c\frac{k}{nD^4}$ for $k\le\sqrt{n}$) and similarly strict rates for Hölder graphons ( $\ge c n^{-(2\gamma+1)/(2\gamma+2)}/D^4$ for $\gamma>0.5$ ), highlighting a fundamental statistical-computational gap. The authors also provide a matching low-degree polynomial upper bound in SBM, achieved by a power-iteration–based estimator followed by a gradient-descent normalization, yielding a near-optimal $\tilde{O}(k/n)$ rate. Connecting graphon estimation with community detection, they derive a lower bound on clustering error that aligns with the generalized Kesten-Stigum threshold and extend the analysis to sparse graphons and biclustering, illustrating the broad reach of the proposed framework and its implications for computational feasibility in network analysis.

Abstract

Graphon estimation has been one of the most fundamental problems in network analysis and has received considerable attention in the past decade. From the statistical perspective, the minimax error rate of graphon estimation has been established by Gao et al (2015) for both stochastic block model and nonparametric graphon estimation. The statistical optimal estimators are based on constrained least squares and have computational complexity exponential in the dimension. From the computational perspective, the best-known polynomial-time estimator is based universal singular value thresholding, but it can only achieve a much slower estimation error rate than the minimax one. The computational optimality of the USVT or the existence of a computational barrier in graphon estimation has been a long-standing open problem. In this work, we provide rigorous evidence for the computational barrier in graphon estimation via low-degree polynomials. Specifically, in SBM graphon estimation, we show that for low-degree polynomial estimators, their estimation error rates cannot be significantly better than that of the USVT under a wide range of parameter regimes and in nonparametric graphon estimation, we show low-degree polynomial estimators achieve estimation error rates strictly slower than the minimax rate. Our results are proved based on the recent development of low-degree polynomials by Schramm and Wein (2022), while we overcome a few key challenges in applying it to the general graphon estimation problem. By leveraging our main results, we also provide a computational lower bound on the clustering error for community detection in SBM with a growing number of communities and this yields a new piece of evidence for the conjectured Kesten-Stigum threshold for efficient community recovery. Finally, we extend our computational lower bounds to sparse graphon estimation and biclustering.

Computational Lower Bounds for Graphon Estimation via Low-degree Polynomials

TL;DR

This work establishes rigorous computational lower bounds for graphon estimation by applying the low-degree polynomial framework to SBM-based priors and nonparametric graphons. It proves that, for degree- polynomial estimators, the estimation risk cannot beat the USVT benchmark up to polylog factors in the SBM regime ( for ) and similarly strict rates for Hölder graphons ( for ), highlighting a fundamental statistical-computational gap. The authors also provide a matching low-degree polynomial upper bound in SBM, achieved by a power-iteration–based estimator followed by a gradient-descent normalization, yielding a near-optimal rate. Connecting graphon estimation with community detection, they derive a lower bound on clustering error that aligns with the generalized Kesten-Stigum threshold and extend the analysis to sparse graphons and biclustering, illustrating the broad reach of the proposed framework and its implications for computational feasibility in network analysis.

Abstract

Graphon estimation has been one of the most fundamental problems in network analysis and has received considerable attention in the past decade. From the statistical perspective, the minimax error rate of graphon estimation has been established by Gao et al (2015) for both stochastic block model and nonparametric graphon estimation. The statistical optimal estimators are based on constrained least squares and have computational complexity exponential in the dimension. From the computational perspective, the best-known polynomial-time estimator is based universal singular value thresholding, but it can only achieve a much slower estimation error rate than the minimax one. The computational optimality of the USVT or the existence of a computational barrier in graphon estimation has been a long-standing open problem. In this work, we provide rigorous evidence for the computational barrier in graphon estimation via low-degree polynomials. Specifically, in SBM graphon estimation, we show that for low-degree polynomial estimators, their estimation error rates cannot be significantly better than that of the USVT under a wide range of parameter regimes and in nonparametric graphon estimation, we show low-degree polynomial estimators achieve estimation error rates strictly slower than the minimax rate. Our results are proved based on the recent development of low-degree polynomials by Schramm and Wein (2022), while we overcome a few key challenges in applying it to the general graphon estimation problem. By leveraging our main results, we also provide a computational lower bound on the clustering error for community detection in SBM with a growing number of communities and this yields a new piece of evidence for the conjectured Kesten-Stigum threshold for efficient community recovery. Finally, we extend our computational lower bounds to sparse graphon estimation and biclustering.
Paper Structure (42 sections, 25 theorems, 154 equations, 1 algorithm)

This paper contains 42 sections, 25 theorems, 154 equations, 1 algorithm.

Table of Contents

  1. Introduction
  2. Our Contributions
  3. From Community Detection to Graphon Estimation
  4. Related Prior Work
  5. Statistical-computational Trade-offs
  6. Why the Low-degree Polynomial Framework
  7. More Literature on Low-degree Polynomials
  8. Organization of the Paper
  9. Notation and Preliminaries
  10. Computational Lower Bounds for Estimation via Low-degree Polynomials
  11. Computational Limits for Graphon Estimation in Stochastic Block Model
  12. Optimality of Theorem \ref{['th: comp-limit-graphon-est']}
  13. A Matching Low-degree Polynomial Upper Bound for $\mathbb{P}_{{\textnormal{SBM}} (p,q)}$
  14. Computational Limits for Nonparametric Graphon Estimation
  15. Computational Limits for Community Detection in SBM
  16. ...and 27 more sections

Key Result

Theorem 1

Suppose $2 \leq k \leq \sqrt{n}$. For any $D \geq 1$, there exists a universal constant $c > 0$ such that Here the notation $\widehat{M} \in \mathbb{R}[A]^{n \times n}_{\leq D}$ means that for all $(i,j) \in [n] \times [n]$, $\widehat{M}_{ij}$ is a polynomial of $A$ with degree no more than $D$.

Theorems & Definitions (41)

  • Theorem 1
  • Proposition 1: schramm2020computational
  • Theorem 2
  • Corollary 1
  • Theorem 3
  • Remark 1
  • Theorem 4
  • Theorem 5
  • Theorem 6
  • Proposition 2
  • ...and 31 more