Learning Graphs With Monotone Topology Properties and Multiple Connected Components

@article{Pavez2017LearningGW,
  title={Learning Graphs With Monotone Topology Properties and Multiple Connected Components},
  author={Eduardo Pavez and Hilmi E. Egilmez and Antonio Ortega},
  journal={IEEE Transactions on Signal Processing},
  year={2017},
  volume={66},
  pages={2399-2413}
}
Recent papers have formulated the problem of learning graphs from data as an inverse covariance estimation problem with graph Laplacian constraints. While such problems are convex, existing methods cannot guarantee that solutions will have specific graph topology properties (e.g., being a tree), which are desirable for some applications. The problem of learning a graph with topology properties is in general non-convex. In this paper, we propose an approach to solve these problems by decomposing… 
View on IEEE
doi.org
44 Citations
Highly Influential Citations
3
Background Citations
27
Methods Citations
15
Results Citations
1

Figures and Tables from this paper

44 Citations

Closed Form Solutions of Combinatorial Graph Laplacian Estimation under Acyclic Topology Constraints

This paper shows that when the target graph topology does not contain any cycle, then the solution has a closed form in terms of the empirical covariance matrix, which enables us to efficiently construct a tree graph from data, even if there is only a single data sample available.

ON LEARNING LAPLACIANS OF TREE STRUCTURED GRAPHS

This paper shows that when the target graph topology is known and does not contain any cycle, i.e., it is a tree, then the optimal Laplacian has a closed form in terms of the empirical covariance matrix.

Learning graphs with monotone topology properties

This work proposes a tractable algorithm that finds the generalized Laplacian matrix of a graph with the desired type of topology by solving a combinatorial optimization problem to find a graph topology that satisfies the desired structural property.

A Unified Framework for Structured Graph Learning via Spectral Constraints

This paper introduces a unified graph learning framework lying at the integration of Gaussian graphical models and spectral graph theory, and develops an optimization framework that leverages graph learning with specific structures via spectral constraints on graph matrices.

Optimization Algorithms for Graph Laplacian Estimation via ADMM and MM

Numerical experiments show that the inclusion of a nominal eigensubspace significantly improves the estimation of the graph Laplacian, which is more evident when the sample size is smaller than or comparable to the problem dimension.

Joint Network Topology Inference via Structured Fusion Regularization

A general graph estimator based on a novel structured fusion regularization that enables us to jointly learn multiple graph Laplacian matrices with such complex topological patterns, and enjoys both high computational efficiency and rigorous theoretical guarantee is proposed.

Nonconvex Sparse Graph Learning under Laplacian Constrained Graphical Model

This paper proposes a nonconvex penalized maximum likelihood estimation method, and establishes the order of the statistical error, and proves that a large regularization parameter will surprisingly lead to a solution representing a complete graph.

On Sparse Graph Estimation Under Statistical and Laplacian Constraints

  • Jitendra Tugnait
  • Computer Science
    2021 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC)
  • 2021
This work augments the objective function of Kalofolias (2016) which is motivated by a signal smoothness viewpoint and imposes a Laplacian constraint, with a penalized log-likelihood objective function with a lasso constraint, motivated from a statistical viewpoint.

Bayesian Topology Learning and noise removal from network data

A graph topology inference approach is proposed to learn the underlying graph structure from a given set of noisy multi-variate observations, which are modeled as graph signals generated from a Gaussian Markov Random Field (GMRF) process.

Graph Learning From Filtered Signals: Graph System and Diffusion Kernel Identification

A novel graph signal processing framework for building graph-based models from classes of filtered signals that reduces to a graph Laplacian estimation problem and demonstrates that the proposed algorithm outperforms the current state-of-the-art methods.

References

SHOWING 1-10 OF 69 REFERENCES

A connectedness constraint for learning sparse graphs

This article shows how connectedness of undirected graphs can be formulated as an analytical property and can be enforced as a convex constraint and especially shows how the constraint relates to the distributed consensus problem and graph Laplacian learning.

Graph Learning from Data under Structural and Laplacian Constraints

This paper proposes a novel framework for learning/estimating graphs from data and demonstrates that the proposed algorithms outperform the current state-of-the-art methods in terms of graph learning performance.

Variants of the Graph Laplacian with Applications in Machine Learning

This thesis contributes a novel re-interpretation of the smallest eigenvalue of the signed Laplacian as the canonical candidate for a spectral approach to correlation clustering and implements the suggested algorithm, shown to compete with state-of-the-art.

The f-Adjusted Graph Laplacian: a Diagonal Modification with a Geometric Interpretation

The f -adjusted graph is introduced and it is proved that it provides the correct cuts and volumes as the sample size tends to infinity and represents a natural family of diagonal perturbations of the original normalized Laplacian.

Generalized Laplacian precision matrix estimation for graph signal processing

  • Eduardo PavezAntonio Ortega
  • Computer Science
    2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
  • 2016
The graph learning problem is formulated as a precision matrix estimation with generalized Laplacian constraints, and a new optimization algorithm is proposed that is more sparse than the ones obtained using other graph learning methods.

Exact Covariance Thresholding into Connected Components for Large-Scale Graphical Lasso

For a range of values of λ, this proposal splits a large graphical lasso problem into smaller tractable problems, making it possible to solve an otherwise infeasible large-scale problem.

Equivalence of graphical lasso and thresholding for sparse graphs

  • S. Sojoudi
  • Computer Science
    2014 52nd Annual Allerton Conference on Communication, Control, and Computing (Allerton)
  • 2014
A simple condition under which the computationally-expensive graphical lasso behaves the same as the simple heuristic method of thresholding is derived, which depends only on the solution of graphicalLasso and makes no direct use of the sample correlation matrix or the regularization coefficient.

Learning A Structured Optimal Bipartite Graph for Co-Clustering

A novel co-clustering method to learn a bipartite graph with exactly k connected components, where k is the number of clusters, which approximates the original graph but maintains an explicit cluster structure.

Spectral sparsification of graphs: theory and algorithms

It is explained what it means for one graph to be a spectral approximation of another and the development of algorithms for spectral sparsification are reviewed, including a faster algorithm for finding approximate maximum flows and minimum cuts in an undirected network.

A Divide-and-Conquer Method for Sparse Inverse Covariance Estimation

This paper considers the regularized sparse inverse covariance matrix estimation problem with a very large number of variables, and proposes a clustering algorithm that attempts to minimize this bound, and in practice, is able to find effective partitions of the variables.
...