一种参数化模糊联想记忆网络的鲁棒性分析

Robustness Analysis of Parameterized Fuzzy Associative Memory Network

基于最大运算Max以及带参数ξ的t-模Tξ的模糊关系合成,提出一种参数化的广义模糊联想记忆网络Max-TξFAM及一种有效学习算法。由于Tξ中参数ξ的作用,在应用中Max-TξFAM有更大的适应性和灵活性。从理论上证明采用该学习算法时,对任意ξ∈[0,1],Max-TξFAM对训练模式摄动的鲁棒性差。通过一个图像联想方面的实验检验了该结论的正确性。

Based on fuzzy composition of maximum operation and a t-norm Tξ with a parameter ξ,a parameterized general fuzzy associative memory network Max-Tξ FAM and its effective learning algorithm are presented.By adjusting parameter ξ,the Max-Tξ FAM has good adaptability and flexibility in practice.It is proved theoretically that,using the mentioned above learning algorithm,Max-Tξ holds weak robustness to perturbations of training pattern pairs for any ξ ∈[0.1].Experiment about image association validates this conclusion.