RobustLDS: Learning linear dynamical systems under adversarial corruptions
Abstract
We consider the problem of learning linear dynamical systems under adversarial contamination from a single trajectory of length T. While identification of linear dynamical systems itself is well-studied, the problem of robust system identification under adversarial contamination is relatively less explored. In this work, we study the setting where a fraction of the T observations are contaminated by adversarial outliers. We propose different estimators based on relaxations of least-trimmed squares along with an alternating minimization algorithm. Furthermore, we also propose two estimators which exploit the group-sparsity (through penalization/hard-constraints) of the outliers. For the estimator with group-sparse penalty, we derive non-asymptotic error bounds which establish its robustness to outliers. We also show empirically that the proposed estimators work well in practice.
Related
- Nearly Minimax Optimal Regret for Adversarial Linear Bandits with Strong Corruptions
- Byzantine-tolerant distributed learning of finite mixture models under partial corruptions
- Symmetric Hermite quadrature-based balanced truncation for learning linear dynamical systems from derivative data
- When Clean Data Hurts: Learning with Monotone Corruptions Beyond Binary Classification
- Model reduction in time: Learning flow maps for reducing stochastic dynamical systems in time
BibTeX
@misc{ruwanpathirana2026robustlds,
title = {{RobustLDS: Learning linear dynamical systems under adversarial corruptions}},
author = {Aravinda Kanchana Ruwanpathirana and Hemant Tyagi},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.11906},
eprint = {2610.11906},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.11906},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - RobustLDS: Learning linear dynamical systems under adversarial corruptions
AU - Aravinda Kanchana Ruwanpathirana
AU - Hemant Tyagi
PY - 2026
DA - 2026/10/08
PB - arXiv
DO - 10.48550/arxiv.2610.11906
UR - https://arxiv.org/abs/2610.11906
AB - We consider the problem of learning linear dynamical systems under adversarial contamination from a single trajectory of length T. While identification of linear dynamical systems itself is well-studied, the problem of robust system identification under adversarial contamination is relatively less explored. In this work, we study the setting where a fraction of the T observations are contaminated by adversarial outliers. We propose different estimators based on relaxations of least-trimmed squares along with an alternating minimization algorithm. Furthermore, we also propose two estimators which exploit the group-sparsity (through penalization/hard-constraints) of the outliers. For the estimator with group-sparse penalty, we derive non-asymptotic error bounds which establish its robustness to outliers. We also show empirically that the proposed estimators work well in practice.
ER -