图像分类卷积神经网络的反馈损失计算方法改进
|
摘要
当前在图像分类领域,卷积神经网络主要通过反向传播算法训练权重和偏置.在参数的训练过程中,网络的实际输出与样本标签之间的反馈损失计算方式会影响到卷积神经网络对图像的最终分类性能.本文研究发现,当增大训练样本标签的维度,提高不同类别标签间的最小汉明距离,并通过sigmoid激活函数结合交叉熵计算反馈损失时,所得到的卷积网络模型对图像的分类能力优于使用softmax激活函数结合独热编码计算反馈损失所得到的卷积网络模型的分类能力.本文使用多种卷积神经网络结构,并结合多个数据集进行训练和测试,所得到的仿真结果证明了本文观点的正确性.
In recent years,convolutional neural networks( CNN) in image classification generally update the weights and offsets through the back-propagation algorithm. Therefore,a suitable algorithm to calculate the feed-back loss is pretty important for CNN,which greatly affects the classification accuracy of the CNN. In this paper,an improved loss calculation algorithm is proposed to improve the image classification accuracy of the CNN. The simulation experiments of this paper show that increasing the dimensions and hamming distance of training sample tags has a positive effect on the classification accuracy of CNN. And in terms of the image classification accuracy,the method of using the sigmoid activation function and the crossentropy loss function to calculate the feedback loss performs better than the method of using softmax activation function and the log-likelihood loss function to calculate the feedback loss.
In recent years,convolutional neural networks( CNN) in image classification generally update the weights and offsets through the back-propagation algorithm. Therefore,a suitable algorithm to calculate the feed-back loss is pretty important for CNN,which greatly affects the classification accuracy of the CNN. In this paper,an improved loss calculation algorithm is proposed to improve the image classification accuracy of the CNN. The simulation experiments of this paper show that increasing the dimensions and hamming distance of training sample tags has a positive effect on the classification accuracy of CNN. And in terms of the image classification accuracy,the method of using the sigmoid activation function and the crossentropy loss function to calculate the feedback loss performs better than the method of using softmax activation function and the log-likelihood loss function to calculate the feedback loss.
引文
[1]Lu Hong-tao,Zhang Qin-chuan.A review of the application of deep convolutional neural networks in computer vision[J].Journal of Data Acquisition&Processing,2016,31(1):1-17.
[2]Redmon J,Divvala S,Girshick R,et al.You only look once:unified,real-time object detection[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition,Las Vegas,Nevada:IEEE Computer Society Press,2016:779-788.
[3]Li Xu-dong,Ye Mao,Li Tao.Research on target detection based on convolutional neural network[J].Application Research of Computers,2017,34(10):2881-2886.
[4]Ouyang Gu,Zhong Bi-neng,Bai Bing,et al.Recent research advances and application of object tacking algorithm based on deep neural network[J].Journal of Chinese Computer Systems,2018,39(2):315-323.
[5]Valmadre J,Bertinetto L,Henriques J,et al.End-to-end representation learning for correlation filter based tracking[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition,Honolulu,Hawaii:IEEE Computer Society Press,2017:5000-5008.
[6]Jonathan Long,Evan Shelhamer,Trevor Darrell.Fully convolutional networks for semantic segmentation[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2017,39(4):640-651.
[7]Zou Guo-feng,Fu Gui-xia,Gao Ming-liang,et al.Pose varied face recognition based on self learning deep convolutional neural network[J].Journal of Chinese Computer Systems,2018,39(6):1156-1162.
[8]Lécun Y,Bottou L,Bengio Y,et al.Gradient-based learning applied to document recognition[J].Proceedings of the IEEE,1998,86(11):2278-2324.
[9]Krizhevsky A,Sutskever I,Hinton G E.ImageNet classification with deep convolutional neural networks[C]//Neural Information Processing Systems,Lake Tahoe:MIT Press,2012:1097-1105.
[10]Szegedy C,Liu W,Jia Y,et al.Going deeper with convolutions[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition,Boston:IEEE Computer Society Press,2015:1-9.
[11]He K,Zhang X,Ren S,et al.Deep residual learning for image recognition[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition,Las Vegas:IEEE Computer Society Press,2016:770-778.
[12]Ioffe S,Szegedy C.Batch normalization:accelerating deep network training by reducing internal covariate shift[C]//International Conference on Machine Learning,Lille:Springer-Verlag,2015:448-456.
[13]He K,Zhang X,Ren S,et al.Identity mappings in deep residual networks[C]//European Conference on Computer Vision,Amsterdam:Springer-Verlag,2016:630-645.
[14]Glorot X,Bordes A,Bengio Y.Deep sparse rectifier neural networks[C]//International Conference on Artificial Intelligence and Statistics,Fort Lauderdale:MLR Press,2011:315-323.
[15]He K,Zhang X,Ren S,et al.Delving deep into rectifiers:surpassing human-level performance on imageNet classification[C]//International Conference on Computer Vision,Santiago:IEEE Computer Society Press,2015:1026-1034.
[16]Zagoruyko S,Komodakis N.Wide residual networks[C]//British Machine Vision Conference,York,UK:BMVA Press,2016:946-957.
[17]Rao Qian,Yu Wen,Mao Qi-qi,et al.Overview of error correction output codes[J].Computer Knowledge and Technology,2013,9(6):1423-1426.
[1]卢宏涛,张秦川.深度卷积神经网络在计算机视觉中的应用研究综述[J].数据采集与处理,2016,31(1):1-17.
[3]李旭冬,叶茂,李涛.基于卷积神经网络的目标检测研究综述[J].计算机应用研究,2017,34(10):2881-2886.
[4]欧阳谷,钟必能,白冰,等.深度神经网络在目标跟踪算法中的应用与最新研究进展[J].小型微型计算机系统,2018,39(2):315-323.
[7]邹国锋,傅桂霞,高明亮,等.基于自学习深度卷积神经网络的姿态变化人脸识别[J].小型微型计算机系统,2018,39(6):1156-1162.
[17]饶倩,喻文,毛祺琦,等.纠错输出码综述[J].电脑知识与技术,2013,9(6):1423-1426.
[2]Redmon J,Divvala S,Girshick R,et al.You only look once:unified,real-time object detection[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition,Las Vegas,Nevada:IEEE Computer Society Press,2016:779-788.
[3]Li Xu-dong,Ye Mao,Li Tao.Research on target detection based on convolutional neural network[J].Application Research of Computers,2017,34(10):2881-2886.
[4]Ouyang Gu,Zhong Bi-neng,Bai Bing,et al.Recent research advances and application of object tacking algorithm based on deep neural network[J].Journal of Chinese Computer Systems,2018,39(2):315-323.
[5]Valmadre J,Bertinetto L,Henriques J,et al.End-to-end representation learning for correlation filter based tracking[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition,Honolulu,Hawaii:IEEE Computer Society Press,2017:5000-5008.
[6]Jonathan Long,Evan Shelhamer,Trevor Darrell.Fully convolutional networks for semantic segmentation[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2017,39(4):640-651.
[7]Zou Guo-feng,Fu Gui-xia,Gao Ming-liang,et al.Pose varied face recognition based on self learning deep convolutional neural network[J].Journal of Chinese Computer Systems,2018,39(6):1156-1162.
[8]Lécun Y,Bottou L,Bengio Y,et al.Gradient-based learning applied to document recognition[J].Proceedings of the IEEE,1998,86(11):2278-2324.
[9]Krizhevsky A,Sutskever I,Hinton G E.ImageNet classification with deep convolutional neural networks[C]//Neural Information Processing Systems,Lake Tahoe:MIT Press,2012:1097-1105.
[10]Szegedy C,Liu W,Jia Y,et al.Going deeper with convolutions[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition,Boston:IEEE Computer Society Press,2015:1-9.
[11]He K,Zhang X,Ren S,et al.Deep residual learning for image recognition[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition,Las Vegas:IEEE Computer Society Press,2016:770-778.
[12]Ioffe S,Szegedy C.Batch normalization:accelerating deep network training by reducing internal covariate shift[C]//International Conference on Machine Learning,Lille:Springer-Verlag,2015:448-456.
[13]He K,Zhang X,Ren S,et al.Identity mappings in deep residual networks[C]//European Conference on Computer Vision,Amsterdam:Springer-Verlag,2016:630-645.
[14]Glorot X,Bordes A,Bengio Y.Deep sparse rectifier neural networks[C]//International Conference on Artificial Intelligence and Statistics,Fort Lauderdale:MLR Press,2011:315-323.
[15]He K,Zhang X,Ren S,et al.Delving deep into rectifiers:surpassing human-level performance on imageNet classification[C]//International Conference on Computer Vision,Santiago:IEEE Computer Society Press,2015:1026-1034.
[16]Zagoruyko S,Komodakis N.Wide residual networks[C]//British Machine Vision Conference,York,UK:BMVA Press,2016:946-957.
[17]Rao Qian,Yu Wen,Mao Qi-qi,et al.Overview of error correction output codes[J].Computer Knowledge and Technology,2013,9(6):1423-1426.
[1]卢宏涛,张秦川.深度卷积神经网络在计算机视觉中的应用研究综述[J].数据采集与处理,2016,31(1):1-17.
[3]李旭冬,叶茂,李涛.基于卷积神经网络的目标检测研究综述[J].计算机应用研究,2017,34(10):2881-2886.
[4]欧阳谷,钟必能,白冰,等.深度神经网络在目标跟踪算法中的应用与最新研究进展[J].小型微型计算机系统,2018,39(2):315-323.
[7]邹国锋,傅桂霞,高明亮,等.基于自学习深度卷积神经网络的姿态变化人脸识别[J].小型微型计算机系统,2018,39(6):1156-1162.
[17]饶倩,喻文,毛祺琦,等.纠错输出码综述[J].电脑知识与技术,2013,9(6):1423-1426.