Dramatically valuable survey paper on inverse reinforcement learning. It describes from the basic mechanism to the applications. The typical methods of inverse reinforcement learning (Max Margin/Max Entropy/Bayesian) are well organized and written. arxiv.org/abs/1806.06877 https://twitter.com/icoxfog417/status/1012664138026311680?s=21 https://arxiv.org/pdf/1806.06877.pdf

GAN], inverse reinforcement learning, and energy-based model can be regarded as the same thing, and techniques from other communities can be used, as long as the generated model G is given a likelihood. For example, the autoregressive model can be used to stabilize learning if the current GAN G is given a likelihood. arxiv.org/abs/1611.03852 https://twitter.com/hillbig/status/811454974274060288?s=21 https://arxiv.org/pdf/1806.06877.pdf https://arxiv.org/abs/1611.03852


This page is auto-translated from /nishio/逆強化学習 using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I’m very happy to spread my thought to non-Japanese readers.