Online Graph Learning In Dynamic Environments

Inferring the graph topology that characterizes structured data is pivotal to many graph-based models when predefined graphs are not available. This paper focuses on learning graphs in the case of sequential data in dynamic environments. For sequential data, an online version of classic batch graph learning method is developed. To better track dynamic graphs, we assume the graphs evolve in some certain patterns such that dynamic priors might be embedded in the online graph learning framework. When we have no knowledge of the hidden patterns, a data-driven method is leveraged to predict the evolution of graphs. Furthermore, dynamic regret analysis of the proposed method is performed, illustrating theoretically that proper dynamic priors do reduce regret. Experimental results support that our method is superior to the state-of-the-art methods.