摘要： 在非线性模式识别中,设计了一种利用Stripe算法调整辅助神经元的新方法,该方法利用相互平行的超平面将权值空间进行划分,使模型的非线性模式识别能力大大提高,用Matlab随机产生非线性分类问题,利用新模型得到较好的分类识别效果,证明了网络结构简单,非线性识别能力较高。

关键词: 结构型神经网络, 遗传算法, 主神经元, 副神经元, 多阈值神经元

Abstract: This paper raises a new constructive neural network to solve the no-linear problem of the pattern classification, which is based on the ancillary neuron parameter adjustment of Stripe algorithm. This model uses the parallel hyperplane to incise the weight space of the network, and raises the networks no-linear pattern classification ability. For proving the efficiency of the model, it random raises some no-linear pattern classification problems and uses the new model to solve them, the new model get a wonderful identification result and proves the model's no-linear identification ability.

中图分类号: 

• TP183