Abstract:The eigenvalue problem of a graph Laplacian matrix L arising from a simple, connected and undirected graph has been given more attention due to its extensive applications, such as spectral clustering, community detection, complex network, image processing and so on. The associated matrix L is symmetric, positive semi-definite, and is usually large and sparse. It is often of interest for finding some smallest positive eigenvalues and corresponding eigenvectors.However, the singularity of L makes the classical eigensolvers inefficient since we need to factorize L for the purpose of solving large and sparse linear systems exactly. The next difficulty is that it is usually prohibitive to factorize L generated by real network problems from big data such as social media transactional databases, and sensor systems because there is in general not only local connections.In this paper, we propose a trimming to cure the singularity of L according to its special property: zero row/column sum. This remedy technique leads us to solve a positive definite linear system reduced in one dimension and then recover the result to get a suitable solution of the original system involved in an eigensolver. Besides, we apply a deflating approach to exclude the influence of converged eigenvalues. We show how to apply the idea of trimming to the graph Laplacian eigenvalue problem together with a deflated term and a target shift. Accordingly, based on the inexact residual Arnoldi (Lee, 2007; Lee and Stewart, 2007) method, we propose an integrated eigensolver for this kind of L combining with the implicit remedy of the singularity, an effective deflation for convergent eigenvalues and the shift–invert enhancement.Numerical experiments reveal that the integrated eigensolver outperforms the classical Arnoldi/Lanczos method for computing some smallest positive eigeninformation especially when the LU factorization is not available.

Left-right neural network for singularvalue decompositions of general matrices

Recurrent Neural Network for Computation of Generalized Eigenvalue Problem with Real Diagonalizable Matrix Pair and Its Applications.

Cross-Correlation Neural Network Models For The Smallest Singular Component Of General Matrix

Lie Neurons: Adjoint-Equivariant Neural Networks for Semisimple Lie Algebras

LOCUS: A Novel Decomposition Method for Brain Network Connectivity Matrices using Low-rank Structure with Uniform Sparsity

Modified cross-correlation neural networks

Neural Network Approaches For The Extraction Of The Eigenstructure

A novel neural network approach for computing modulus largest or smallest eigen-pairs of arbitrary real matrices

Orthogonal Constrained Neural Networks for Solving Structured Inverse Eigenvalue Problems

Tensor Neural Network Models for Tensor Singular Value Decompositions.

Neural Network Layer Matrix Decomposition reveals Latent Manifold Encoding and Memory Capacity

Lie Group Decompositions for Equivariant Neural Networks

A Conditional Singular Value Decomposition

A Riemannian Network for SPD Matrix Learning

Isira: Integrated Shift-Invert Residual Arnoldi Method for Graph Laplacian Matrices from Big Data

Matrix Manifold Neural Networks++

A New Neural Network for L(1)-Norm Programming

Riemannian Local Mechanism for SPD Neural Networks

A Neurodynamic Approach to Compute the Generalized Eigenvalues of Symmetric Positive Matrix Pair.

A complex neural network for computing modulus largest eigen-pairs of real antisymmetric matrices

Neural network models for the quaternion singular value decompositions