论文部分内容阅读
基于已经提出的Lukasiewicz t-模算子的模糊双向联想记忆网络(FBAM)的学习算法,进一步研究该网络的性质。在理论上证明了只要存在使给定的模式对集合能成为FMBAM的平衡态集合,则该连接权矩阵对能使FMBAM对任意输入全局收敛到平衡态。当训练模式存在摄动时,利用该学习算法训练的FBAM,对训练模式摄动拥有好的鲁棒性。
The properties of this network are further studied based on the learning algorithm of Fuzzy Bidirectional Associative Memory Network (FBAM) with Lukasiewicz t-modulo operator. It is theoretically proved that this connection matrix matrix can make FMBAM converge to any equilibrium globally to any input as long as there exists a set of equilibrium states that make FMBAM a given set of patterns. When the training pattern is perturbed, the FBAM trained by this learning algorithm has good robustness to training pattern perturbation.