Potts网络的模块化Latching动力模型

Latching Dynamics in Potts Network

Latching(锁连)是意大利国际高等研究生院的Alessandro Treves教授根据人类语言所独有的无穷递归能力假设构造出的自适应Potts动力学模型。Latching能够解释思维中的策略转移现象,对于探索人类智能的出现具有重要的意义。但是Latching动力学模型是建立在单个模块基础上的,无法与实验证据进行直接对话。利用具有时延的异联想突触连接将Latching动力扩展到模块化Latching链,并对其主要参数重绕概率、阈值系数、噪声模式对和反馈连接系数做了深入的仿真分析。研究的主要结果有:(1)一个适当的重绕概率能增强模块内模式之间的锁连活动。(2)在模块化网络中,LCL和ISR对反馈连接和噪声模式的大小都不敏感。(3)只有在阈值较高,网络中才会出现比较清晰的Latching转移。

Latching dynamics, designed by A. Treves from SISSA, is a kind of adaptive Potts dynamics model that derived from the unique infinite recursion capability hypothesis in human languages. It successfully explains the strategic transition phenomenon in human thinking, bringing great significance to the exploration of the emergence of human intelligence. But Latching dynamics runs on a single module, forfeiting a fruitful dialogue with the experimental evidence. A modular Latching chain is proposed by the introduction of delayed hereto - associative synapses, and the effects of structural parameters like rewiring probility, threshold, noise patterns pairs and feedback connections are analyzed by computational simulations. The main findings in the paper includes ： 1 ） an appropriate rewiring probability promotes the intra - modular latching transitions. 2） The average latching chain length and inverse switching ratio are insensitive to the noise and feedback connections. 3） A larger threshold is necessary for the clear transition between modules.