一个仿真突触超微结构的三阶神经网络模型

A 3RD-ORDER NEURAL NETWORK MODEL SIMULATING SYNAPSE ULTRAMICROSTRUCTURE

神经系统内突触类型的多样性是与其机能的复杂性密切相关的.考虑到这一特性.本文用一个三阶神经网络模型仿真突触超微结构中的复合型连续性突触结构.作出了该模型的理论推导和分析.对由40个单元组成的网络系统作了计算机仿真、所得结果表明其存储容量可达680个图样.对其吸引盆也进行了仿真分析,对新模型与传统的Hebb学习律下的Hopfield模型进行了对比仿真.发现新模型具有比后者快得多的演化速率.

The variety of the types of synapses in neural systems has strong relation with the complexity of its function. Considering this. We used a 3rd-order neural network model to simulate the function of the complex dendrospinous in synapse ultramicrostructure . The theory of the model was given and analysed. The results of computer simulation of the system with 40 units showed that the capacity of the network was approximately 680 patterns.Analysis and computer simulation of attraction basins were carried out. Compared calculations of computer simulation between our model and Hopfield's model under traditional Hebb's law showed that the former was much faster in retrieval rate.