人脸识别上的进步 56 ▣ 正确率95.17%[D.Chen,X.Cao,F.WenJ.Sun,CVPR13] 正确率97.35%[Y.Taigman,,M.Yang,M.Ranzato,L.Wolf,CVPR14 ▣ 正确率99.47%[Y.Sun,X.Wang,andX.Tang,CVPR14 ▣ 正确率99.63%F.Schroff,D.Kalenichenko,andJ.Philbin,CVPR15] 在LFW上，2年内错误率从5%下降到0.5% 错300对→错30对) IMAGENET DBN Google BP Science Speech Face Microsoft 1986 2006 20112012 20142015 电子科技大学研究生《机器学习》
电子科技大学研究生《机器学习》 人脸识别上的进步 o 正确率95.17% [D.Chen, X. Cao, F. Wen, J. Sun, CVPR13] o 正确率97.35% [Y.Taigman, M. Yang, M.Ranzato, L. Wolf, CVPR14] o 正确率99.47% [Y. Sun, X. Wang, and X. Tang, CVPR14] o 正确率99.63% [F. Schroff, D. Kalenichenko, and J. Philbin, CVPR15] 16 1986 2006 DBN Science Speech 2011 2012 Face 2014 2015 BP 在LFW上，2年内错误率从5%下降到0.5% (错300对错30对)
多层神经网络 56 C3:f.maps 16@10x10 INPUT C1:feature maps S4:f.maps 16@5x5 32x32 6@28x28 S2:f.maps 6@14x14 C5:layer F6:layer OUTPUT 120 84 10 Full connection Gaussian connections Convolutions Subsampling Convolutions Subsampling Full connection LeNet 5 RESEARCH 000 answer: 0 0 103 17 电子科技大学研究生《机器学习》
电子科技大学研究生《机器学习》 多层神经网络 17
LeNet-5 56 C3:f.maps 16@10x10 INPUT C1:feature maps S4:f.maps 16@5x5 32x32 6@28x28 S2:f.maps 6@14x14 C5:layer F6:layer OUTPUT 120 84 10 Full connection Gaussian connections Convolutions Subsampling Convolutions Subsampling Full connection 口七层网络是如何构造的？ 口如何快速编程实现应用模型开发？ 电子科技大学研究生《机器学习》
电子科技大学研究生《机器学习》 LeNet-5 o 七层网络是如何构造的？ o 如何快速编程实现应用模型开发？
LeNet-5 56 C3:f.maps 16@10x10 INPUT C1:feature maps S4:f.maps 16@5x5 6@28x28 32x32 S2:f.maps 6@14x14 C5:layer F6:layer OUTPUT 120 84 10 Full connection Gaussian connections Convolutions Subsampling Convolutions Subsampling Full connection 口七层结构 C1卷积层 S2池化层 ▣ 三大关键技术 C3 卷积层 局部感受野 S4池化层 权值共享 C5 卷积层 池化 F6 全连接层 ■ F7 全连接层 电子科技大学研究生《机器学习》
电子科技大学研究生《机器学习》 LeNet-5 o 七层结构 n C1 卷积层 n S2 池化层 n C3 卷积层 n S4 池化层 n C5 卷积层 n F6 全连接层 n F7 全连接层 o 三大关键技术 n 局部感受野 n 权值共享 n 池化
卷积神经网络CNN 54 K.Fukushima,"Neocognitron:A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position,"Biological Cybernetics,vol 36,pp.193-202,1980 ■ Y.LeCun,B.Boser,J.S.Denker,D.Henderson,R.E. Howard,W.Hubbard,and L.D.Jackel,"Backpropagation applied to handwritten zip code recognition,"Neural Computation,vol.1,no.4,pp.541-551,1989 ■ Y.Le Cun,L.Bottou,Y.Bengio,and P.Haffner, "Gradient-based learning applied to document recognition," Proceedings of the IEEE,vol.86,no.11,pp.2278-2324, 1998 电子科技大学研究生《机器学习》
电子科技大学研究生《机器学习》 卷积神经网络CNN n K. Fukushima, “Neocognitron: A self-organizing neural network model for a mechanism of pattern recognition unaf ected by shift in position, ” Biological Cybernetics, vol. 36, pp. 193–202, 1980 n Y. LeCun, B. Boser, J. S. Denker, D. Henderson, R. E. Howard, W. Hubbard, and L. D. Jackel, “Backpropagation applied to handwritten zip code recognition, ” Neural Computation, vol. 1, no. 4, pp. 541–551, 1989 n Y. Le Cun, L. Bottou, Y. Bengio, and P. Haffner, “Gradient-based learning applied to document recognition, ” Proceedings of the IEEE, vol. 86, no. 11, pp. 2278–2324, 1998