Policy_Based

pick the best actor I’m showing log probabilities (-1.2, -0.36) for UP and DOWN instead of the raw probabilities (30% and 70% in this case) because we always optimize the log probability of the correc

>>阅读原文<<