摘要
神经网络模式识别方法是模式识别领域的一个新的研究方向,而BP算法是神经网络众多算法中应用最为广泛的一种。针对BP算法易于陷入局部极小值且收敛速度慢的缺陷,在BP神经网络训练过程中采用混合学习策略,并对建立的货币识别模型进行Matlab仿真。实验结果表明:MBP-RO模型使神经网络缩短了训练时间,获得了更高的识别速度和更好的识别效果,该模型在货币识别中具有一定的优势。
Pattern-recognition methods of nerve network are a new research direction in pattern-recognition domain and BP algorithm is the most widespread one in the nerve network algorithms. To overcome the drawbacks of falling into local minimum and slow convergence during Back-Propgation(BP) neural network learning, employed a hybrid learning strategy and MATLAB simulation is made to analyze result of the currency identification modeL The experiment indicates: MBP-RO model neural networks has shortened training time and gained higher speed and better outcomes,it shows the model have advantage in currency identification.
出处
《电脑与信息技术》
2008年第3期29-31,共3页
Computer and Information Technology