Learning Low-Complexity Autoregressive Models via Proximal Alternating Minimization

TL;DR

This paper introduces a novel method using proximal alternating minimization to estimate low-complexity vector autoregressive models by effectively combining limited time sequence data with abundant steady-state data.

Contribution

The paper develops a PALM-based approach for low-complexity autoregressive model estimation with convergence guarantees and explicit proximal formulas, improving efficiency and accuracy.

Findings

PALM converges globally to a critical point.

The method reduces estimation error monotonically.

Outperforms gradient projection in efficiency and solution quality.

Abstract

We consider the estimation of the state transition matrix in vector autoregressive models, when time sequence data is limited but nonsequence steady-state data is abundant. To leverage both sources of data, we formulate the least squares minimization problem regularized by a Lyapunov penalty. We impose cardinality or rank constraints to reduce the complexity of the autoregressive model. We solve the resulting nonconvex, nonsmooth problem by using the proximal alternating linearization method (PALM). We show that PALM is globally convergent to a critical point and that the estimation error monotonically decreases. Furthermore, we obtain explicit formulas for the proximal operators to facilitate the implementation of PALM. We demonstrate the effectiveness of the developed method on synthetic and real-world data. Our experiments show that PALM outperforms the gradient projection method in both computational efficiency and solution quality.