Research and application of dynamic neural network based on reinforcement learning

Abstract

Dynamic neural network is became one of the most important approaches to eliminate Q table (look-up-table) of the machine intelligence. On the basis of comparison between general Artificial neural network and dynamic neural network, the development of dynamic neural network will be discussed. After the introduction of the theory and algorithms of reinforcement learning (RL), dynamic neural network will be applied as a basic decision taking unit (classifier neural network) in the form of a new technology. This will develop the application of reinforcement learning and provides a new idea for agent learning during real time operation. Use neural network for supervised learning, state as input/action as label. Reinforcement learning is widely use by different research field as intelligent control, robotics and neuroscience. It provides us possible solution within unknown environment. But at the same time we have to take care of its decision because RL can independently learn without prior knowledge or training and it take decision by learning experience through trail-and-error interaction with its environment. In this paper, we discussed a new dynamic neural network model and its algorithms in detail, together with the issues that arise in Q table (look-up-table).Additionally, the benefit and challenges of reinforcement learning are described along with some of the problem domains where the dynamic neural network techniques have been applied. In order to access dynamic neural network is to eliminate Q table (look-up-table) and agent should learn during real time operation. © 2012 Springer-Verlag GmbH Berlin Heidelberg.