Reinforcement learning has a wide applying prospect in automatic drive and game control and becomes hotter and hotter these years. We start to immersed in these fields and expect to achieve more goals in application. Reinforcement learning is a kind of technology of machine learning. In this model, we set four parameters: environment, reward, action and state. When machine take “action” to achieve the “reward”, we set a parameter to stand for “environment”. If the operations of machine tend to that parameter, we give “state” to it so as to strengthen this behavior, and vice versa. In this process, we simulate natural selection to choose the advantageous selection for us. Human obtains breakthough by learning from the nature once again. With vast operations the control results can be achieved as expected and this is “reinforcement”.
References 《机器学习盛宴——记ICML 2018》