Sample Efficient Algorithms for Linear System Identification under Noisy Observations

In this paper, we focus on learning linear dynamical systems under noisy observations. In this setting, existing algorithms either yield biased parameter estimates, or suffer from large sample complexities. To address these issues, we adapt the instrumental variable method and the bias compensation method, originally proposed for error-in-variables models, to our setting and provide refined non-asymptotic analysis. Under mild conditions, our algorithms achieve superior sample complexities that match the best-known sample complexity for learning a fully observable system without observation noise.