English

Efficient and passive learning of networked dynamical systems driven by non-white exogenous inputs

Machine Learning 2022-05-09 v3 Artificial Intelligence

Abstract

We consider a networked linear dynamical system with pp agents/nodes. We study the problem of learning the underlying graph of interactions/dependencies from observations of the nodal trajectories over a time-interval TT. We present a regularized non-casual consistent estimator for this problem and analyze its sample complexity over two regimes: (a) where the interval TT consists of nn i.i.d. observation windows of length T/nT/n (restart and record), and (b) where TT is one continuous observation window (consecutive). Using the theory of MM-estimators, we show that the estimator recovers the underlying interactions, in either regime, in a time-interval that is logarithmic in the system size pp. To the best of our knowledge, this is the first work to analyze the sample complexity of learning linear dynamical systems \emph{driven by unobserved not-white wide-sense stationary (WSS) inputs}.

Keywords

Cite

@article{arxiv.2110.00852,
  title  = {Efficient and passive learning of networked dynamical systems driven by non-white exogenous inputs},
  author = {Harish Doddi and Deepjyoti Deka and Saurav Talukdar and Murti Salapaka},
  journal= {arXiv preprint arXiv:2110.00852},
  year   = {2022}
}

Comments

Accepted in AISTATS 2022

R2 v1 2026-06-24T06:34:40.801Z