A0968
Title: A regularized low tubal-rank model for high-dimensional time series data
Authors: Samrat Roy - Indian Institute of Management Ahmedabad (India) [presenting]
Abstract: High-dimensional time series analysis has diverse applications in macroeconomics and finance. Recent factor-type models employing tensor-based decompositions prove to be computationally involved due to the non-convex nature of the underlying optimization problem, and they do not capture the underlying temporal dependence of the latent factor structure. The concept of tubal rank is leveraged, and a matrix-valued time series model is developed, which first captures the temporal dependence in the data, and then the remainder signals across the time points are decomposed into two components: a low tubal rank tensor representing the baseline signals, and a sparse tensor capturing the additional idiosyncrasies in the signal. The issue of identifiability of various components is addressed in the model, and a scalable alternating block minimization algorithm is subsequently developed to solve the convex regularized optimization problem for estimating the parameters. Finite sample error bounds are provided under high dimensional scaling for the model parameters.
