Graph Matching via convex relaxation to the simplex

Ernesto Araya Valdivia,Hemant Tyagi

DOI: https://doi.org/10.3934/fods.2024034

2024-08-09

Abstract:This paper addresses the Graph Matching problem, which consists of finding the best possible alignment between two input graphs, and has many applications in computer vision, network deanonymization and protein alignment. A common approach to tackle this problem is through convex relaxations of the NP-hard \emph{Quadratic Assignment Problem} (QAP). Here, we introduce a new convex relaxation onto the unit simplex and develop an efficient mirror descent scheme with closed-form iterations for solving this problem. Under the correlated Gaussian Wigner model, we show that the simplex relaxation admits a unique solution with high probability. In the noiseless case, this is shown to imply exact recovery of the ground truth permutation. Additionally, we establish a novel sufficiency condition for the input matrix in standard greedy rounding methods, which is less restrictive than the commonly used `diagonal dominance' condition. We use this condition to show exact one-step recovery of the ground truth (holding almost surely) via the mirror descent scheme, in the noiseless setting. We also use this condition to obtain significantly improved conditions for the GRAMPA algorithm [Fan et al. 2019] in the noiseless setting.

Machine Learning,Statistics Theory

What problem does this paper attempt to address?

### Problems Addressed by the Paper

The paper primarily addresses the Graph Matching problem. Specifically, the Graph Matching problem refers to finding the optimal alignment between two input graphs. This problem has numerous applications in fields such as computer vision, network de-anonymization, and protein structure alignment.

The paper proposes a novel convex relaxation method and develops an efficient mirror descent algorithm to solve this problem. Under the related Gaussian Wigner model, the authors prove that this convex relaxation method has a unique solution and can accurately recover the true permutation in the absence of noise. Additionally, the paper introduces a new sufficient condition for the input matrix in the standard greedy rounding method, which is more relaxed than the commonly used "diagonal dominance" condition. With this condition, the paper demonstrates that the mirror descent algorithm can achieve one-step accurate recovery of the true permutation in the absence of noise. Furthermore, this condition also improves the conditions for the GRAMPA algorithm in the absence of noise.

The research methods were evaluated on both synthetic and real data, showing that their statistical performance is superior to existing convex relaxation methods, with similar computational costs.