Spectral Filtering for General Linear Dynamical Systems

Abstract

We give a polynomial-time algorithm for learning latent-state linear dynamical systems without system identification, and without assumptions on the spectral radius of the system's transition matrix. The algorithm extends the recently introduced technique of spectral filtering, previously applied only to systems with a symmetric transition matrix, using a novel convex relaxation to allow for the efficient identification of phases.

References in corpus (2)

• On the Sample Complexity of the Linear Quadratic Regulator
• Online Learning for Time Series Prediction

Cited by in corpus (11)

• On the Sample Complexity of the Linear Quadratic Regulator
• Learning Without Mixing: Towards A Sharp Analysis of Linear System Identification
• Active Learning for Nonlinear System Identification with Guarantees
• Regularized Variational Data Assimilation for Bias Treatment using the Wasserstein Metric
• No-Regret Prediction in Marginally Stable Systems
• Fairness in Forecasting of Observations of Linear Dynamical Systems
• Online Adaptive Principal Component Analysis and Its extensions
• PAC-Bayes Generalisation Bounds for Dynamical Systems Including Stable RNNs
• Robust guarantees for learning an autoregressive filter
• Global Convergence Using Policy Gradient Methods for Model-free Markovian Jump Linear Quadratic Control
• Online Convex Optimization in Changing Environments and its Application to Resource Allocation