Deep Reinforcement Learning: From Q-Learning to Deep Q-Learning

As the two hottest branches of machine learning, deep learning and reinforcement learning both play a vital role in the field of artificial intelligence. Combining deep learning with reinforcement learning, deep reinforcement learning is a method of artificial intelligence that is much closer to human learning. As one of the most basic algorithms for reinforcement learning, Q-learning is a discrete strategic learning algorithm that uses a reasonable strategy to generate an action. According to the rewards and the next state generated by the interaction of the action and the environment, optimal Q-function can be obtained. Furthermore, based on Q-learning and convolutional neural networks, the deep Q-learning with experience replay is developed in this paper. To ensure the convergence of value function, a discount factor is involved in the value function. The temporal difference method is introduced to training the Q-function or value function. At last, a detailed procedure is proposed to implement deep reinforcement learning.