论文部分内容阅读
借鉴仿生学原理,基于大脑皮层结构提出一种新型侧抑制神经网络(S-LINN)模型.通过模拟大脑皮层内锥体神经元和抑制神经元的连接特点,在多层结构的S-LINN的不同层神经元之间引入跨越连接,同时在隐含层内神经元之间进行信息的侧向抑制传输.引入的两种连接机制有效地提高了网络处理问题的能力,与其他网络相比能够以更精简的结构较好地解决实际问题.通过对乳腺癌诊断数据集和异或问题的求解,表明了S-LINN网络不但能够获得较高的训练精度,而且具有更强的泛化能力.
Based on the theory of bionics, a novel lateral inhibitory neural network (S-LINN) model was proposed based on the cerebral cortex structure. By simulating the pyramidal neurons and inhibiting neuronal connections in the cerebral cortex, Cross-connection between different layers of neurons is introduced, and lateral suppression of transmission of information between neurons in the hidden layer is also introduced.The two connection mechanisms introduced effectively improve the ability of the network to deal with the problem, compared with other networks The solution to the problem of breast cancer diagnosis datasets and XOR problems shows that the S-LINN network can not only obtain higher training accuracy, but also have more generalization ability.