论文部分内容阅读
当前神经网络是重新激励人们去探索的一个课题。这种模拟脑功能的信息存储和运算系统,非常适合复杂问题的并行处理。现在,在实验室已研制了一些应用装置,在社会上也有了许多工业应用的商品。网络拓扑学、学习算法、及用VLSI(very Large scale Integrated circuit,超大规模集成电路)实施的电路系统等三个研究领域的新进展正在孕育着神经计算机(Neuro computer)的诞生。一、神经计算机的特点离散符号模式的数字机在统治全球,而连接论(connectionism)模式的感知机受冷落近20载。然而,情况在变,现在让亿次级克雷计算机来识别哪是树木,哪是电线杆,它却黔驴技穷。人工智能领域的人们似乎要重新搬出连接论方法,以适应复杂问题的解
The current neural network is to re-inspire people to explore a subject. This information storage and computing system that mimics brain function is ideal for parallel processing of complex problems. Now, some applications have been developed in the laboratory, and there are also many industrial applications in the community. New advances in three areas of research, including network topology, learning algorithms, and circuitry implemented with VLSI (Very Large Scale Integrated Circuits), are nurturing the birth of a Neuro computer. First, the characteristics of the Nerve Computer Discrete symbolic mode of the number of machines in the rule of the world, and the connection theory (connectionism) moderator of the machine was cold for nearly 20 years. However, the situation is changing, so now to make billions of times Cray computer to identify which is the tree, which is the power pole, it is poor technology. People in the field of artificial intelligence seem to have to reintroduce the connective approach to the solution of complex problems