分子计算理论与神经网络

本文以“智能—问题求解—信息处理—学习和记忆—系统实现”为线索,介绍了分子计算及神经网络的基本概念和原理,提出了它们的研究层次及其异同点,详细讨论了进化学习算法和神经网络的自组织原则,并利用最大信息准则,给出了无导师学习算法的统一描述框图.本文最后讨论了分子计算与神经网络的实现问题,认为分子电子器件是神经网络实现的一条可能途径.

Molecnlar Compating and Neural Networks

In this paper,the basic concepts and principles of molecular compu- ting and netural network are introduced.Their research hierarchies are presen- ted and compared with each other.The evolutionary learning algorithm and self-organizing principle of neural networks are discussed.Based on the ma- ximum-information rule,a diagram integratedly describing the unsupervised learning algorithm is proposed.Finally,problems on implementing molecular computing systems and neural networks are discussed,and the use of molecu- lar electronic devices is suggested as a promising approach to implement neural networks.