一种基于GA-BP学习的神经网络推理机制的研究  

Study on GA-BP-based neural network inference mechanism

作　　者：雷宇[1] 李孝安[1] 

机构地区：[1]西北工业大学计算机学院,陕西西安710129

出　　处：《电子设计工程》2013年第12期34-36,39,共4页Electronic Design Engineering

摘　　要：为了克服基于规则专家系统推理适应性差、学习能力差等缺点,本文对基于GA算法与BP算法的神经网络专家系统推理进行了研究,给出了一种基于GA-BP学习的神经网络推理模型及实现算法。使用UCI标准数据集进行仿真实验,同其他典型的四种推理算法进行了比较研究,得出基于GA-BP模型的推理具有学习速度快,实时性和准确性高的优点。In order to overcome the disadvantages in inference adaptability of the traditional rule-based expert system, we investigate the inference mechanism for multi-layer neural networks based on genetic algorithm and back-propagation algorithm. A GA-BP based algorithm for neural network inference is presented and compared with other four kinds of algorithms. The preliminary experimental result for the UCI data sets shows that the reasoning model based on GA-BP algorithm demonstrates better performance with faster learning speed, higher accuracy and adaptability.

关 键 词：专家系统 推理 神经网络 BP算法 遗传算法 

分 类 号：TP389[自动化与计算机技术—计算机系统结构]