摘要
在文献[1]提出的模糊形态联想记忆网络FMAM的基础上,提出了一种新型的模糊形态学双向联想记忆网络FMBAM,证明了FMBAM的双向联想中能够保证记忆在一步之内完成,因此不存在收敛问题,也表明了FM-BAM具有优越的抗腐蚀或膨胀噪声的能力。但是,通常的噪声是随机的,为此,本文提出了动态核的方法,从而较好地提高了FMBAM对随机噪声的抗噪能力。仿真实验验证了利用动态核的双向联想记忆网络FMBAM。在加入较大的随机噪声的情况下,仍能保证完全记忆在一步之内完成。
Based on fuzzy morphological associative memory (FMAM)[1], the new bi-directional fuzzy morphological associative memory (FMBAM) is presented in this paper. FMBAM avoids the so-called convergence problem, due to the fact that it can reach perfect recall within one step. Another advantage is its strong robustness in suppressing erosion or dilation noise. In order to effectively suppress random noise rather than erosion or dilation noise in real datasets, the dynamic kernel approach is proposed to enhance FMBAM's robustness for random noise. Our experimental results here demonstrate the effectiveness of the dynamic kernel approach.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2005年第3期257-262,共6页
Pattern Recognition and Artificial Intelligence
基金
国家自然科学基金(No.60225015)
江苏省自然科学基金(No.BK2003017)
关键词
动态核
随机噪声
模糊形态学
双向联想记忆
模糊神经网络
Dynamic Kernel
Random Noise
Fuzzy Morphology
Bi - Directional Associative Memory
Fuzzy Neural Networks