Turing Equivalence of Fuzzy Max-Min Operator Neural Networks
-
摘要: 将模糊Zadeh算子的定义域作了扩充,并重新定义为模糊极大极小算子,使其满足交换律、结合律和零元律.在此基础上提出一种模糊极大极小算子型神经元网络模型,符合一般模糊算子型神经元网络的定义.与传统的Zadeh算子型模糊神经网络相比具有较强的映射能力.详细证明了用该模糊极大极小算子神经元网络可以计算与图灵机等价的部分递归函数,从而表明模糊极大极小算子神经元网络的计算能力等价于图灵机.将传统神经元M-P模型神经网络的图灵等价性推广到模糊神经元网络.Abstract: The definition region of Zadeh fuzzy operator is extended and the max-min operator is redefined such that it satisfies the exchange law, the combination law and the 0-element law. On the above basis, a max-min operator neural network is proposed according with the general definition of fuzzy operator neural networks. Comparing with traditional fuzzy Zadeh operator neural networks, the present network has high mapping ability. It is showed in detail that the max-min operator neural network can compute part-recursion function, which is equivalent to Turing machine. This indicates that fuzzy max-min operator neural network has the same computation ability as Turing machine. This extends the result of Turing equivalence of traditional neural networks of neuron M-P model to fuzzy neural networks.
-
Key words:
- fuzzy operators /
- neural networks /
- recursive functions /
- Turing equivalence
-
[1] 梁久祯,何新贵. 单体模糊神经网络的函数逼近能力[J]. 计算机研究与发展, 2000, 37(9):1045~1049. [2] 刘晓鸿, 戴汝为. 线性阈值单元神经元网络的图灵等价性[J]. 计算机学报, 1995, 18(6):438~442 . [3] 李晓忠, 汪培庄, 罗承忠. 模糊神经网络[M]. 贵阳:贵州科技出版社, 1994. [4] Cutland N. Computability:an introduction to recursive function theory . Cambridge:Cambridge Univ Press, 1980.
点击查看大图
计量
- 文章访问数: 2944
- HTML全文浏览量: 71
- PDF下载量: 1053
- 被引次数: 0