基于AP和BP神经网络算法的手写数字识别

his paper proposes a method of handwritten digit recognition combined AP with BP neural network algorithm to recognise large-scale digital handwriting quickly and efficiently. AP algorithm is firstly uesd to cluster training samples to eliminate redundant and re-construct the sample space, then BP neural network is utilized to learn and recognise each class from AP clustering. Experiments were conducted with the data from UCI machine learning database，and results show that the correct identification rate of this method is no less than that of BP neural network algorithm and that the processing time is only its tenth. Thus, this method can be used to handwritten digits quickly and efficiently identify with high practical value.