aiqbench analysis
This paper advances robust system identification by proposing estimators capable of learning linear dynamical systems despite adversarial contamination in observation trajectories. By leveraging group-sparsity constraints and alternating minimization, the authors derive non-asymptotic error bounds that empirically validate resilience against outliers. The work addresses a critical gap in ensuring model reliability when training data is compromised, which is essential for deploying AI agents in untrusted environments.
RobustLDS: Learning linear dynamical systems under adversarial corruptions
arXiv:2610.11906v1 Announce Type: cross 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.
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.
Machine Learning (stat.ML); Machine Learning (cs.LG); Systems and Control (eess.SY); Optimization and Control (math.OC); Statistics Theory (math.ST)
arXiv:2610.11906 [stat.ML]
(or arXiv:2610.11906v1 [stat.ML] for this version)
https://doi.org/10.48550/arXiv.2610.11906 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 228 words
Related coverage
Robust Multi-Agent Reinforcement Learning for Small UAS Separation Assurance under GPS Degradation and Spoofing
arXiv · Aug 25, 2026
Lattice-Based Cryptography Explained
Quantum Zeitgeist · Aug 15, 2026
Online Control via Counterfactual Tracking
arXiv · Oct 6, 2026
L-Lipschitz Gershgorin ResNet Network
arXiv · Sep 16, 2026