基于卷积神经网络的手写数字识别研究

Handwritten Digit Recognition Based on Convolution Neural Network

为了解决传统的多层神经网络在手写体数字识别中网络参数过多、计算量大、准确率低等问题,将卷积神经网络应用于手写体数字识别,设计了一种新的网络模型,在MNIST数据集上对网络模型进行训练;对比分析常用的优化算法在不同学习率时对识别效果的影响,选用具有自适应学习率的优化算法.利用MNIST数据集进行测试,本文模型识别准确率可以达到99%以上,损失接近于0.实验结果表明,利用该网络模型进行手写数字识别,识别率高、运算量小,提高了识别速度和准确率,具有很强的鲁棒性.

In order to solve the problems of too many network parameters,large amount of calculation and low accuracy of traditional multilayer neural network in handwritten digit recognition,a new network model is designed and trained on the MINIST dataset by using convolution neural network in handwritten digit recognition.By comparing and analyzing the influence of common optimization algorithms on recognition performance at different learning rates,an optimization algorithm with adaptive learning rate is selected.Using the MNIST data set,the recognition accuracy of this model can reach more than 99%and the loss is close to 0.Experiments show that by using this network model for handwritten digit recognition,the accuracy is high,the computation is small,the recognition speed and accuracy are improved,and it has strong robustness.