在本篇論文中，我們提出一個O(1)時間的類神經網路，以將n個資料排序
，其所需的類神經元及連結個數為O(n^(1+?), ?> 0。就解決大型的問
題而言，我們需要快速的硬體方案。為了能在常數時間內使用較少的硬體
成本來達成排序任務，我們採用了Leighton 的column sort 作為主要的
架構，並以Chen和Hsieh的類神經網路 作為最底層的子網路。 經過適當
的安排，我們得以發展出常數時間而且成本較低的類神經排序網路。本篇
論文 之結果已於一九九五年十月為國際性期刊NeuroComputing所接受，
並將出版。

Neural network is a proper model for parallel computing because it can process the data simultaneously. Based on the model of neural network, many problems can be done in O(1) time. These problems include summation, weighted summation, WTA(winner-take-all) problem, sorting, and so on. However, most researches use lots of neurons to achieve the O(1) time complexity. For large-size problems, it is desirable to have low-cost hardware solutions. In this paper, we will present an O(1) time neural network with O(n^(1+?) neurons and links to sort n data, ?> 0. In order to solve the sorting problem in constant time and with less hardware-cost, we adopt Leighton's column sort as the main architecture. Then we use Chen and Hsieh's neural network with O(n^2) neurons and links as the lowest-level subnetworks.By using recursive technique properly, we are able to explore constant-time, low-cost neural sorting networks.