摘要
针对传统数字字符识别算法收敛速度慢且有可能陷入局部极小值等问题,提出了将BFGS拟牛顿算法应用于含噪数字字符识别:构造前馈型神经网络,调用Matlab神经网络工具箱中的训练函数trainbfg对网络进行训练.该算法收敛速度快、识别精度高,能够对含有一定噪声的数字字符进行识别,具有广阔的应用前景.
The traditional numeric character recognition algorithm which has slow convergence speed and might fall into the local minimum.To solve such problems,the BFGS quasi-Newton algorithm was presented that was applied to the recognition of numeric characters.First,a feed-forward neural network was set up,then network was trained by calling trainbfg on Matlab.The algorithm has high accuracy,fast convergence,can recognize the numeric characters with noise efficiently,so it has broad application prospects.
出处
《郑州轻工业学院学报（自然科学版）》
CAS
2011年第4期79-81,共3页
Journal of Zhengzhou University of Light Industry:Natural Science
