Sparse plus low-rank autoregressive identification in neuroimaging time\n series

2015/03/30 by Raphaël Liégeois, Liégeois, Raphaël, Bamdev Mishra +5
Computer Science · Mathematics · Engineering · #Blind Source Separation Techniques #Statistical and numerical algorithms #Sparse and Compressive Sensing Techniques

paper · pdf · doi:10.48550/arxiv.1503.08639

Abstract

This paper considers the problem of identifying multivariate autoregressive\n(AR) sparse plus low-rank graphical models. Based on the corresponding problem\nformulation recently presented, we use the alternating direction method of\nmultipliers (ADMM) to efficiently solve it and scale it to sizes encountered in\nneuroimaging applications. We apply this decomposition on synthetic and real\nneuroimaging datasets with a specific focus on the information encoded in the\nlow-rank structure of our model. In particular, we illustrate that this\ninformation captures the spatio-temporal structure of the original data,\ngeneralizing classical component analysis approaches.\n

Related

• Dimensionality reduction for time series data
• Time-varying Autoregression with Low Rank Tensors
• Sparse plus low-rank identification for dynamical latent-variable graphical AR models
• AR Identification of Latent-variable Graphical Models
• A Bayesian Approach to Sparse plus Low rank Network Identification