Yu F. Richard, He Ying. Deep Reinforcement Learning for Wireless Networks

Springer, 2019. — 120 р.There is a phenomenal burst of research activities in machine learning and wireless systems. Machine learning evolved from a collection of powerful techniques in AI areas and has been extensively used in data mining, which allows the system to learn the useful structural patterns and models from training data. Reinforcement learning is an important branch of machine learning, where an agent learns to take actions that would yield the most reward by interacting with the environment. The main advantage of reinforcement learning is that it works well without prior knowledge of an exact mathematical model of the environment. However, the traditional reinforcement learning approach has some shortcomings, such as low convergence rate to the optimal behavior policy and its inability to solve problems with high-dimensional state space and action space. These shortcomings can be addressed by deep reinforcement learning.The key idea of deep reinforcement learning is to approximate the value function by leveraging the powerful function approximation property of deep neural networks. After training the deep neural networks, given a state-action pair as input, deep reinforcement learning is able to estimate the long-term reward. The estimation result can guide the agent to choose the best action.