Learning Mixtures of Linear Dynamical Systems

2022/01/26 by Yanxi Chen, H. Vincent Poor, Chen, Yanxi +1 · 3 citations
Engineering · Computer Science · #Control Systems and Identification #Gaussian Processes and Bayesian Inference #Machine Learning and Algorithms

paper · pdf · doi:10.48550/arxiv.2201.11211

Abstract

We study the problem of learning a mixture of multiple linear dynamical systems (LDSs) from unlabeled short sample trajectories, each generated by one of the LDS models. Despite the wide applicability of mixture models for time-series data, learning algorithms that come with end-to-end performance guarantees are largely absent from existing literature. There are multiple sources of technical challenges, including but not limited to (1) the presence of latent variables (i.e. the unknown labels of trajectories); (2) the possibility that the sample trajectories might have lengths much smaller than the dimension d of the LDS models; and (3) the complicated temporal dependence inherent to time-series data. To tackle these challenges, we develop a two-stage meta-algorithm, which is guaranteed to efficiently recover each ground-truth LDS model up to error O(√(d/T)), where T is the total sample size. We validate our theoretical studies with numerical experiments, confirming the efficacy of the proposed algorithm.

Cited by

• Learning Mixtures of Linear Dynamical Systems via Hybrid Tensor-EM Method
• Conformal Prediction for Time-series Forecasting with Change Points
• Learning clusters of partially observed linear dynamical systems

Related

• Tensor Decompositions Meet Control Theory: Learning General Mixtures of Linear Dynamical Systems
• Finite Sample Analysis of Tensor Decomposition for Learning Mixtures of Linear Systems
• Learning clusters of partially observed linear dynamical systems
• Joint Problems in Learning Multiple Dynamical Systems
• Learning Mixtures of Markov Chains and MDPs