论文部分内容阅读
鉴于学习样本对神经网络模型的模式识别性能有很大的影响 ,提出学习样本的选择应与识别模型所利用的特性相结合 ,并利用汉明 (Hamming)距离对用于旋转不变识别的级联模型的学习样本进行优选 ,计算机对三个很相似的飞机模型进行识别 ,识别结果表明对学习样本进行有效的选择不仅可以减少系统的学习训练时间而且可以提高模型的识别能力。
In view of the great influence of learning samples on the performance of neural network model recognition, this paper proposes that the choice of learning samples should be combined with the characteristics used by the recognition model and the Hamming distance is used to detect the rotation invariant The results show that the effective selection of learning samples can not only reduce the learning training time of the system but also improve the recognition ability of the model.