论文部分内容阅读
回声状态网络是近年来新兴的一种递归神经网络,独特而简单的训练方式以及高精度的训练结果已使其成为当前研究的热点之一.在该网络中,引入了储备池计算模式这一新的神经网络的建设方案,克服了之前网络模型基于梯度下降的学习算法所难以避免的收敛慢和容易陷入局部极小等问题.围绕这种新型网络结构,国内外许多学者开展了多样的研究.本文全面深入介绍了回声状态网络这一新兴技术,讨论了回声状态网络的优缺点,并综合近年的研究现状,总结了回声状态网络的主要研究工作进展和未来的研究方向.
Echo state network is a kind of recursive neural network which is emerging in recent years. The unique and simple training method and high-precision training result have made it become one of the hot spots in the current research.In this network, the reserve pool calculation mode The new neural network construction scheme overcomes the problems that the previous network model based on the gradient descent learning algorithm is difficult to avoid the convergence slow and easy to fall into the local minimum.Surrounding this new network structure, many scholars at home and abroad have carried out a variety of studies In this paper, we introduce the new state - of - the - art echo state network in depth and discuss the advantages and disadvantages of the echo state network. Based on the research status in recent years, the progress of the research on the echo state network and the research direction in the future are summarized.