模块化神经网络MATLAB仿真系统的研制

Development of Modular Neural Networks′ MATLAB Simulation System

分析了模块化神经网络MATLAB仿真系统的设计和实现。该仿真系统充分运用MATLAB对数值计算、图形处理的强大功能,采用可扩展的图形用户界面,开发了一套适合模块化神经网络试验和研究平台。介绍了模块化神经网络(MNN)的子网动态集成方法,并提出了一种基于“一专多能”思想的集成训练方法。该仿真系统将算法、数据、仿真演示有机集成,为后期的设计开发提供了有力的支持。

It analyses the design and realization of Modular Neural Networks＇ （MNN） MATLAB simulation system. This system is developed by MATLAB that has powerful functions on data computing and graphics disposal. By using extensible GUI, this simulation system develops a platform for MNN＇s experiment and research. Subnet dynamic integration method of MNN is introduced. And a new subnet training method of MNN based on thinking of ＂Knowing something of everything and everything of something＂ is presented. This system is fully integrated with algorithm, data and simulated display, which provides advantageous support for the advanced design for the later development.