Dual Representations for Dynamic Programming and Reinforcement Learning | IEEE Conference Publication | IEEE Xplore