神经网络识别非平衡样本模式的网络优化方法

Improved Neural-Network Optimization Algorithms for Classification of Imbalanced-Exemplar Patterns

本文基于多目标代化的思想，分析了用神经网络实现非平衡样本模式识别与分类时，网络优化过程中基本BP算法收敛速度低的原因，给出了两种相应的改进算法，并利用这两种算法研究了两例非平衡样本模式的分类问题．研究结果表明：改进的算法有效地提高了网络优化的收敛速度。

Based on the ideas of the multiobjechve optindzation, the pape analyzes the cause that convergence-rate of the standard BP(Back-Propagation) algorithm is low in the optindzaion of netal network which is used for classilication of imbalanced-exemplar patterns. We present the improved optindzation algorithms. Using the algorithms,we have been able to accelerate the rate of learning for two kinds of dsanced-exemplarpattem classification problems. The results indicate that the impmved algorithms can efficiently increase the convereence-rate of neural network optimizahon.