A Convergence analysis of Algorithm

In this section, we provide a convergence rate analysis for Algorithm 1. Similar to Hazan et al. We first give the following proposition that captures certain properties of the proposed objective. Taylor's theorem, one has kr L Source code is included in the supplemental material. MADE's bonus is set to the following in tabular setting: 1 p N Below, we describe details on each tabular algorithm. We implement discounted value iteration given in [37] with all three bonuses.