Reinforcement Learning of Sequential Price Mechanisms

We introduce the use of reinforcement learning for indirect mechanisms, working with existing class sequential price which generalizes both serial dictatorship and posted mechanisms essentially characterizes all strongly obviously strategyproof mechanisms. Learning an optimal mechanism within this forms a partially-observable Markov decision process. provide rigorous conditions when is more powerful than simpler static sufficiency or insufficiency observation statistics learning, necessity complex (deep) policies. show that our approach can learn near-optimal in several experimental settings.