扩展随机神经网络及其概率结构特性分析

Extended random neural network model and its probability structure features

在Ｇｅｌｅｎｂｅ随机神经网络（ＧＮＮ）模型基础上，提出扩展ＧＮＮ（ＥＧＮＮ）模型。考虑信号释放强度依赖于神经元兴奋水平的情形，给出了ＥＧＮＮ的平稳分布。讨论了平稳分布存在的条件。在较弱假设下，ＥＧＮＮ仍具有简洁的“积”形式平稳分布，比原ＧＮＮ增强了调节平稳分布概率结构特性的功能，且能够表达更多的智能和生物特性。例举了它的联想功能。

This paper proposes an extended random neural network (EGNN) model based on the model of Gelenbe's random neural network (GNN). It considers the case that the time intervals between successive signal emissions of a neuron are dependent on the neuron potential, and analyzes the stationary distribution of the EGNN. The paper proves that the EGNN has a stationary distribution of product form which leads to simple expressions for the system state, and shows that it has enhanced functions of adjusting the probability structure comparing with the GNN. Thus, the EGNN is competent in representing many intelligent and biological features (e.g., the feature of associated memory).