摘要
讨论了一种新的并行排序算法．基于前馈阈值神经网络结构，该排序模型利用Ｏ（ｍｎ￣２）个神经元经６个时间步（６级前馈）即可完成排序。
By considering the specific features of parallel computation of neural networks(NNs),anew parallel sorting model based on feedforward threshold NNs is proposed.The basic pro-cessing unit,i.e.,the neuron in the NNs,is a linear threshold device. The selection ofweights is limited to the polynomial level of the input parameters and the weights in each lay-er are fixed. Analysis and simulation show that the sorting could be completed in six timesteps using O(mn￣2)neurons,where n is the number of data to be sorted and m is the binarybit number of each datum.The sorting time is independent of the sorting size. The simplestructure of the neuron makes the NNs suitable for realization with VLSI.
出处
《华中理工大学学报》
CSCD
北大核心
1994年第6期124-128,共5页
Journal of Huazhong University of Science and Technology
关键词
线性阈值函数
并行排序算法
sorting
linear threshold function
feedforward threshold neural network
