论文部分内容阅读
本文是“神经网络与神经计算机的基本原理和应用”的第二部分。主要介绍Hopfleld 网络模型及其在求解旅行商问题中的成功应用。为了避免陷入局部极小,Boltz-mann 机和随机网络模型对 Hopfield 模型作了进一步的改进。感知机和 BP 网络,是利用神经网络进行自学习的两种常见方式。按误差反向传播的 BP层次型网络,可以通过样本训练改变神经元的连接权以进行学习和记忆。在神经计算机一节中, 扼要介绍了神经计算机的主要特征、体系结构、其中并行拉度、虚拟处理器和拓扑结构是三个需要考虑的重要参数。文中最后部分讨论了神经网络与神经计算机的应用领域。
This article is the second part of “Basic Principles and Applications of Neural Networks and Neural Computers.” Introduce Hopfleld network model and its successful application in solving traveling salesman problems. In order to avoid falling into local minimum, the Hopfield model is further improved by Boltzmann machine and stochastic network model. Perception machines and BP networks are two common ways of learning by using neural networks. Backpropagation error-BP hierarchical network, you can use the sample training to change the connection rights of neurons for learning and memory. In the Nerve Computer section, the main features and architecture of the Nerve Computer are briefly introduced. The parallelism, virtual processor and topology are three important parameters to consider. The last part of the article discusses the application fields of neural networks and neural computers.