摘要
回顾了面向神经网络应用的程序设计环境的发展。神经网络应用程序可分为三类：着重应用的，着重算法的和通用的系统。着重应用的系统主要针对市场需要。着重算法的系统支持确定的神经网络模型，它包括两个主要子类：支持单一神经网络模型的系统和支持多种模型算法的系统。第三类是通用编程系统，它提供的“工具箱”包含多种算法和编程工具。编程系统可细分为：用于某一算法和应用的通用系统；用户可调整系统任一部份的开放系统；用来对确定机器编程（例如并行计算机）的着重硬件的系统。我们将讨论多种现有的系统实例和它们的发展。
This paper reviews the development of software environments for programming to neural network applications.We introduce a taxonomy of three classes:application oriented、algerithm oriented and general programming systems.Application oriented systems are designed for market requirement;Algerithm oriented systems support specific neural network models;general programming systems provide tool boxs comprising many algerithms and programming tools.We discuss examples of the numerous systems currently available and their taxonomy.
基金
国家"九五"重点攻关项目资助
