1: \begin{abstract}

2: %   In this paper, we study what characteristics does a data-driven

3: %   feedback controller should posses in order to be robust to

4: %   uncertainties in the experimental data. In particular, we assume a

5: %   linear system being stabilized by data-driven static feedback

6: %   controller, and assume that the measurements are corrupted by random

7: %   noise or a malicious adversary, in order to destabilize the

8: %   system. As the corrupted feedback controller results in a random

9: %   linear system, our results are stated in terms of probability of

10: %   instability. Using some recent advances in random matrix theory, we

11: %   show that (i) the rate of convergence of probability of instability

12: %   towards one is in the order of square root of number of perturbed

13: %   data points, (ii) the controller can ensure stability with respect

14: %   to sparse perturbations in the experimental data, provided it does

15: %   not depend too much on all the data.

16: % \end{abstract}

17: