RobustLDS: Learning linear dynamical systems under adversarial corruptions

Key points

• While identification of linear dynamical systems itself is well-studied, the problem of robust system identification under adversarial contamination is relatively less explored.1
• In this work, we study the setting where a fraction of the $T$ observations are contaminated by adversarial outliers.1
• We propose different estimators based on relaxations of least-trimmed squares along with an alternating minimization algorithm.1
• For the estimator with group-sparse penalty, we derive non-asymptotic error bounds which establish its robustness to outliers.1

Sources (1)

• [1]RobustLDS: Learning linear dynamical systems under adversarial corruptions
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 8, 01:08 PM

  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.

Extractive summary: sentences quoted from the sources.