二维时延有序反向传播神经网络模型与算法

The architecture and algorithm of two--dimensional time delay ordered backpropagation

基于有序系统的概念,文中提出了一种二维时延有序反向传播神经网络模型并给出了相应的有序BP算法,同时指出传统的BP算法可看作是该算法的一种特例。该模型可用于模式识别、图像处理等领域。文中最后对5类飞机模型纵向像识别的实验比较,结果表明该模型及算法比传统的BP算法在性能上有一定的优越性。

Based on the idea of an ordered system, a kind of Two--Dimensional Time Delay Ordered Backpropagation Neural Network (2--D TDOBP) is proposed with the presentation of an or- dered BP algorithm. It is pointed out that the conventional BP algorithm can be regarded as a special case of the ordered BP algorithm. This model can be used in the area of pattern recogni- tion, image processing, etc. Finally the result of a simulation experiment, which is used to distinguish between one--dimensional range images of five kinds of plane models, indicates that, the model and algorithm proposed in this paper is, to certain degree, superior to the con- ventional BP algorithm.