Supplementary Material Contextual Games: Multi-Agent Learning with Side Information Pier Giuseppe Sessa, Ilija Bogunovic, Andreas Krause, Maryam Kamgarpour (NeurIPS 2020)

The theoretical guarantees obtained in Section 3 rely on the following two main lemmas. 's be the distributions computed using the MW rule: p Similarly to Appendix A.1, we let As it was done in proof of Theorems 1 and Appendix A.1, Also, we now explicitly consider the adaptiveness of the adversary. The second equality follows by the law of total expectation. Consider a contextual game and assume contexts are sampled i.i.d. Hoeffding's inequality [21] shows that for any > 0 P E In this section we describe the experimental setup of the contextual traffic routing game of Section 5.