Online Joint Topology Identification and Signal Estimation with Inexact Proximal Online Gradient Descent

Identifying the topology that underlies a set of time series is useful for tasks such as prediction, denoising, and data completion. Vector autoregressive (VAR) model based topologies capture dependencies among time series, and are often inferred from observed spatio-temporal data. When the data are affected by noise and/or missing samples, the tasks of topology identification and signal recovery (reconstruction) have to be performed jointly. Additional challenges arise when i) the underlying topology is time-varying, ii) data become available sequentially, and iii) no delay is tolerated. To overcome these challenges, this paper proposes two online algorithms to estimate the VAR model-based topologies. The proposed algorithms have constant complexity per iteration, which makes them interesting for big data scenarios. They also enjoy complementary merits in terms of complexity and performance. A performance guarantee is derived for one of the algorithms in the form of a dynamic regret bound. Numerical tests are also presented, showcasing the ability of the proposed algorithms to track the time-varying topologies with missing data in an online fashion.