Gradient Descent Learns Linear Dynamical Systems

zbMath1461.62150arXiv1609.05191MaRDI QIDQ4558171

Moritz Hardt, Benjamin Recht, Tengyu Ma

Publication date: 21 November 2018

Full work available at URL: https://arxiv.org/abs/1609.05191

zbMATH Keywords

time seriesnonconvex optimizationlinear dynamical systemstochastic gradient descentgeneralization boundsover-parameterization

Mathematics Subject Classification ID

Linear regression; mixed models (62J05) Non-Markovian processes: estimation (62M09) Inference from stochastic processes and spectral analysis (62M15) Nonconvex programming, global optimization (90C26) Time series analysis of dynamical systems (37M10)

Related Items

Metric entropy limits on recurrent neural network learning of linear dynamical systems, Sequential Subspace Optimization for Quasar-Convex Optimization Problems with Inexact Gradient, On the optimization landscape of tensor decompositions, Data-dependent analysis of model validation errors for linear system identification, Unnamed Item, On the sample complexity of the linear quadratic regulator, Accelerated methods for weakly-quasi-convex optimization problems, First-Order Methods for Nonconvex Quadratic Minimization, Unnamed Item, Unnamed Item, Unnamed Item, A modular analysis of adaptive (non-)convex optimization: optimism, composite objectives, variance reduction, and variational bounds, Unnamed Item

Cites Work

• Probability in Banach spaces. Isoperimetry and processes
• Uniqueness of estimated k-step prediction models of ARMA processes
• Iterative minimization of \(H_2\) control performance criteria
• Some properties of the output error method
• Uniqueness of prediction error estimates of multivariable moving average models
• Robust convergence of the steepest descent method for data-based control
• Atomic Norm Denoising With Applications to Line Spectral Estimation
• A Learning Theory Approach to System Identification and Stochastic Adaptive Control
• Finite sample properties of system identification methods
• Inequalities of A. Markoff and S. Bernstein for polynomials and related functions
• Unnamed Item
• Unnamed Item
• Unnamed Item
• Unnamed Item
• Unnamed Item