A Scalable Strategy for the Identification of Latent-variable Graphical\n Models

2018/09/05 by Daniele Alpago, Alpago, Daniele, Mattia Zorzi +3
Engineering · Computer Science · #Control Systems and Identification #Bayesian Modeling and Causal Inference #Gaussian Processes and Bayesian Inference

paper · pdf · doi:10.48550/arxiv.1809.01608

Abstract

In this paper we propose an identification method for latent-variable\ngraphical models associated to autoregressive (AR) Gaussian stationary\nprocesses. The identification procedure exploits the approximation of AR\nprocesses through stationary reciprocal processes thus benefiting of the\nnumerical advantages of dealing with block-circulant matrices. These advantages\nbecome more and more significant as the order of the process gets large. We\nshow how the identification can be cast in a regularized convex program and we\npresent numerical examples that compares the performances of the proposed\nmethod with the existing ones.\n

Related

• AR Identification of Latent-variable Graphical Models
• Identification of Sparse Reciprocal Graphical Models
• Sparse plus low-rank identification for dynamical latent-variable graphical AR models
• Empirical Bayesian Learning in AR Graphical Models
• Learning Vector Autoregressive Models with Latent Processes