Time-varying Graph Learning Under Structured Temporal Priors

TL;DR

This paper introduces a novel framework for learning time-varying graphs using structured temporal priors, capturing complex temporal relations beyond simple chain models, and employs ADMM for efficient optimization.

Contribution

It proposes a new temporal graph structure to model complex temporal relations and develops an ADMM-based distributed algorithm for learning such graphs.

Findings

Outperforms existing methods in numerical experiments.

Effectively models complex temporal relations.

Demonstrates superior accuracy in graph learning tasks.

Abstract

This paper endeavors to learn time-varying graphs by using structured temporal priors that assume underlying relations between arbitrary two graphs in the graph sequence. Different from many existing chain structure based methods in which the priors like temporal homogeneity can only describe the variations of two consecutive graphs, we propose a structure named \emph{temporal graph} to characterize the underlying real temporal relations. Under this framework, the chain structure is actually a special case of our temporal graph. We further proposed Alternating Direction Method of Multipliers (ADMM), a distributed algorithm, to solve the induced optimization problem. Numerical experiments demonstrate the superiorities of our method.