Efficient Signed Graph Sampling via Balancing & Gershgorin Disc Perfect Alignment

被引:0
作者
Dinesh, Chinthaka [1 ]
Cheung, Gene [2 ]
Bagheri, Saghar [2 ]
Bajic, Ivan V. [3 ]
机构
[1] Northeastern Univ, Coll Engn, Vancouver, BC V6B 1Z3, Canada
[2] York Univ, Dept Elect Engn & Comp Sci, Toronto, ON M3J 1P3, Canada
[3] Simon Fraser Univ, Sch Engn Sci, Burnaby, BC V5A 1S6, Canada
基金
加拿大自然科学与工程研究理事会;
关键词
Laplace equations; Covariance matrices; Sampling methods; Symmetric matrices; Kernel; Eigenvalues and eigenfunctions; Sparse matrices; Interpolation; Electronic mail; Correlation; Graph signal processing; graph spectrum; graph sampling; Gershgorin circle theorem; SET SELECTION; SIGNALS; RECONSTRUCTION; REGULARIZATION; IMAGE;
D O I
10.1109/TPAMI.2024.3524180
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
A basic premise in graph signal processing (GSP) is that a graph encoding pairwise (anti-)correlations of the targeted signal as edge weights is leveraged for graph filtering. Existing fast graph sampling schemes are designed and tested only for positive graphs describing positive correlations. However, there are many real-world datasets exhibiting strong anti-correlations, and thus a suitable model is a signed graph, containing both positive and negative edge weights. In this paper, we propose the first linear-time method for sampling signed graphs, centered on the concept of balanced signed graphs. Specifically, given an empirical covariance data matrix (C) over bar, we first learn a sparse inverse matrix L, interpreted as a graph Laplacian corresponding to a signed graph G. We approximate G with a balanced signed graph G(b) via fast edge weight augmentation in linear time, where the eigenpairs of Laplacian L-b for G(b) are graph frequencies. Next, we select a node subset for sampling to minimize the error of the signal interpolated from samples in two steps. We first align all Gershgorin disc left-ends of LaplacianL(b) at the smallest eigenvalue lambda(min) (L-b) via similarity transform L-s = SLb S-1, leveraging a recent linear algebra theorem called Gershgorin disc perfect alignment (GDPA). We then perform sampling on L-s using a previous fast Gershgorin disc alignment sampling (GDAS) scheme. Experiments show that our signed graph sampling method outperformed fast sampling schemes designed for positive graphs on various datasets with anti-correlations.
引用
收藏
页码:2330 / 2348
页数:19
相关论文
共 65 条
[1]   Efficient Sampling Set Selection for Bandlimited Graph Signals Using Graph Spectral Proxies [J].
Anis, Aamir ;
Gadde, Akshay ;
Ortega, Antonio .
IEEE TRANSACTIONS ON SIGNAL PROCESSING, 2016, 64 (14) :3775-3789
[2]  
[Anonymous], 2004, Gersgorin and His Circles
[3]   Fast Graph Sampling Set Selection Using Gershgorin Disc Alignment [J].
Bai, Yuanchao ;
Wang, Fen ;
Cheung, Gene ;
Nakatsukasa, Yuji ;
Gao, Wen .
IEEE TRANSACTIONS ON SIGNAL PROCESSING, 2020, 68 :2419-2434
[4]  
Bai YC, 2019, INT CONF ACOUST SPEE, P5396, DOI [10.1109/icassp.2019.8682200, 10.1109/ICASSP.2019.8682200]
[5]   Graph-Based Blind Image Deblurring From a Single Photograph [J].
Bai, Yuanchao ;
Cheung, Gene ;
Liu, Xianming ;
Gao, Wen .
IEEE TRANSACTIONS ON IMAGE PROCESSING, 2019, 28 (03) :1404-1418
[6]  
Beckenbach F. E., 1961, Tech. Rep.
[7]  
Biyikoglu T, 2005, ELECTRON J LINEAR AL, V13, P344
[8]   A Constrained l1 Minimization Approach to Sparse Precision Matrix Estimation [J].
Cai, Tony ;
Liu, Weidong ;
Luo, Xi .
JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION, 2011, 106 (494) :594-607
[9]   Greedy Sampling of Graph Signals [J].
Chamon, Luiz F. O. ;
Ribeiro, Alejandro .
IEEE TRANSACTIONS ON SIGNAL PROCESSING, 2018, 66 (01) :34-47
[10]   Manifold Graph Signal Restoration Using Gradient Graph Laplacian Regularizer [J].
Chen, Fei ;
Cheung, Gene ;
Zhang, Xue .
IEEE TRANSACTIONS ON SIGNAL PROCESSING, 2024, 72 :744-761