Equivalence between policy gradients and soft Q-learning

Read on openai.com