Tight Risk Bound for High Dimensional Time Series Completion

Abstract

Initially designed for independent datas, low-rank matrix completion was successfully applied in many domains to the reconstruction of partially observed high-dimensional time series. However, there is a lack of theory to support the application of these methods to dependent datas. In this paper, we propose a general model for multivariate, partially observed time series. We show that the least-square method with a rank penalty leads to reconstruction error of the same order as for independent datas. Moreover, when the time series has some additional properties such as periodicity or smoothness, the rate can actually be faster than in the independent case.

arXiv comment25 pages, 4 figures

References in corpus (14)

• Recovering low-rank matrices from few coefficients in any basis
• Matrix Completion Methods for Causal Panel Data Models
• Restricted strong convexity and weighted matrix completion: Optimal bounds with noise
• Matrix Completion from Noisy Entries
• Noisy low-rank matrix completion with general sampling distribution
• Low Rank and Structured Modeling of High-dimensional Vector Autoregressions
• Factor modeling for high-dimensional time series: Inference for the number of factors
• 1-bit Matrix Completion: PAC-Bayesian Analysis of a Variational Approximation
• Structured low-rank matrix completion for forecasting in time series analysis
• Minimal penalties and the slope heuristics: a survey
• A Bayesian Approach for Noisy Matrix Completion: Optimal Rate under General Sampling Distribution
• Probabilistic low-rank matrix completion on finite alphabets
• High dimensional VAR with low rank transition
• Matrix factorization for multivariate time series analysis