论文部分内容阅读
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。
Reviewed the development of programming environment for neural network applications. Neural network applications fall into three categories: application focused, algorithmic and general purpose systems. Emphasis application of the system mainly for market needs. The algorithm-focused system supports a defined neural network model that includes two major subclasses: a system that supports a single neural network model, and a system that supports multiple model algorithms. The third category is a common programming system, which provides a “toolbox” contains a variety of algorithms and programming tools. The programming system can be subdivided into: a general purpose system for an algorithm and application; an open system for which the user can adjust any part of the system; and a hardware-intensive system for programming a machine, such as a parallel computer. We will discuss a variety of existing system examples and their development.