Verified Safe Reinforcement Learning for Neural Network Dynamic Models | Read Paper on Bytez