Spectral State Compression of Markov Processes

被引:0
|
作者
Zhang, Anru [1 ]
Wang, Mengdi [2 ,3 ]
机构
[1] Univ Wisconsin Madison, Dept Stat, Madison, WI 53706 USA
[2] Princeton Univ, Dept Elect Engn, Princeton, NJ 08540 USA
[3] Princeton Univ, Ctr Stat & Machine Learning, Princeton, NJ 08540 USA
基金
美国国家科学基金会;
关键词
Markov processes; Upper bound; Estimation; Complexity theory; Matrix decomposition; Control theory; Numerical models; Computational complexity; maximum likelihood estimation; minimax techniques; signal denoising; tensor SVD; MATRIX; AGGREGATION; ALGORITHM;
D O I
10.1109/TIT.2019.2956737
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Model reduction of Markov processes is a basic problem in modeling state-transition systems. Motivated by the state aggregation approach rooted in control theory, we study the statistical state compression of a discrete-state Markov chain from empirical trajectories. Through the lens of spectral decomposition, we study the rank and features of Markov processes, as well as properties like representability, aggregability, and lumpability. We develop spectral methods for estimating the transition matrix of a low-rank Markov model, estimating the leading subspace spanned by Markov features, and recovering latent structures like state aggregation and lumpable partition of the state space. We prove statistical upper bounds for the estimation errors and nearly matching minimax lower bounds. Numerical studies are performed on synthetic data and a dataset of New York City taxi trips.
引用
收藏
页码:3202 / 3231
页数:30
相关论文
共 50 条