Sparse plus low-rank autoregressive identification in neuroimaging time series
This work addresses modeling challenges in neuroimaging data analysis, but it is incremental as it builds on a recently presented formulation.
The paper tackles the problem of identifying multivariate autoregressive sparse plus low-rank graphical models in neuroimaging time series, using ADMM to efficiently solve and scale it, and shows that the low-rank structure captures spatio-temporal information, generalizing classical component analysis.
This paper considers the problem of identifying multivariate autoregressive (AR) sparse plus low-rank graphical models. Based on the corresponding problem formulation recently presented, we use the alternating direction method of multipliers (ADMM) to efficiently solve it and scale it to sizes encountered in neuroimaging applications. We apply this decomposition on synthetic and real neuroimaging datasets with a specific focus on the information encoded in the low-rank structure of our model. In particular, we illustrate that this information captures the spatio-temporal structure of the original data, generalizing classical component analysis approaches.