1232 工程科学学报，第43卷，第9期 4ppl,2012,39(3):2947 applied to document recognition.Proc /EEE,1998,86(11):2278 [11]Girshick R.Fast R-CNN /2015 IEEE International Conference [20]Huang G B.Zhu Q Y,Siew C K.Extreme leaming machine: on Computer Vision (ICCV).Santiago,2015:1440 Theory and applications.Neurocomputing,2006,70(1-3):489 [12]Huang G,Liu Z,Van Der Maaten L,et al.Densely connected [21]Huang G B.Chen L,Siew C K.Universal approximation using convolutional networks 2017 IEEE Conference on Computer incremental constructive feedforward networks with random Vision and Pattern Recognition (CVPR).Honolulu,2017:2261 hidden nodes.IEEE Trans Neural Netw,2006,17(4):879 [13]He K M.Zhang X Y,Ren S Q,et al.Spatial pyramid pooling in [22]Anthimopoulos M,Christodoulidis S,Ebner L,et al.Lung pattem deep convolutional networks for visual recognition.IEEE Trans classification for interstitial lung diseases using a deep Pattern Anal Mach Intell,2015.37(9):1904 convolutional neural network.IEEE Trans Med Imaging,2016, [14]Szegedy C,Liu W,Jia Y Q,et al.Going deeper with convolutions 35(5):1207 Il 2015 IEEE Conference on Computer Vision and Pattern [23]Li J Y,Zhao Y K.Xue Z E,et al.A survey of model compression Recognition (CVPR).Boston,2015:1 for deep neural networks.Chin J Eng,2019,41(10):1229 [15]Schroff F,Kalenichenko D,Philbin J.FaceNet:A unified (李江昀，赵义凯，薛卓尔，等.深度神经网络模型压缩综述，工 embedding for face recognition and clustering /2015 /EEE 程科学学报，2019,41(10)：1229) Conference on Computer Vision and Pattern Recognition (CVPR). [24]Ji S W,Xu W,Yang M,et al.3D convolutional neural networks Boston,2015:815 [16]Ding C X,Tao D C.Trunk-branch ensemble convolutional neural for human action recognition.IEEE Trans Pattern Anal Mach networks for video-based face recognition.IEEE Trans Pattern lmel,2013,35(1):221 Anal Mach Intell,2018,40(4):1002 [25]Bartlett P L.The sample complexity of pattern classification with [17]He R,Wu X,Sun Z N,et al.Wasserstein CNN:Leaming invariant neural networks:The size of the weights is more important than features for NIR-VIS face recognition.IEEE Trans Pattern Anal the size of the network.IEEE Trans Inf Theory,1998,44(2):525 Mach Intell,2019,41(7):1761 [26]PhysioNet.MIT-BIH Arrhythmias Database [J/OL].PhysioNet [1]Shelhamer E,Long J,Darrell T.Fully convolutional networks for Online(2020-04-15)[2021-01-12]https://physionet.org/files/mitdb/ semantic segmentation.IEEE Trans Pattern Anal Mach Intell, 1.0.0 2017,39(4):640 [27]Moody G B,Mark R G.The impact of the mit-bih arrhythmia [19]LeCun Y,Bottou L,Bengio Y,et al.Gradient-based learing database.IEEE Eng Med Biol Mag,2001,20(3):45Appl, 2012, 39（3）: 2947 Girshick R. Fast R−CNN // 2015 IEEE International Conference on Computer Vision (ICCV). Santiago, 2015: 1440 [11] Huang  G,  Liu  Z,  Van  Der  Maaten  L,  et  al.  Densely  connected convolutional  networks  //  2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu, 2017: 2261 [12] He K M, Zhang X Y, Ren S Q, et al. Spatial pyramid pooling in deep  convolutional  networks  for  visual  recognition. IEEE Trans Pattern Anal Mach Intell, 2015, 37（9）: 1904 [13] Szegedy C, Liu W, Jia Y Q, et al. Going deeper with convolutions //  2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Boston, 2015: 1 [14] Schroff  F,  Kalenichenko  D,  Philbin  J.  FaceNet:  A  unified embedding  for  face  recognition  and  clustering  //  2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Boston, 2015: 815 [15] Ding C X, Tao D C. Trunk-branch ensemble convolutional neural networks  for  video-based  face  recognition. IEEE Trans Pattern Anal Mach Intell, 2018, 40（4）: 1002 [16] He R, Wu X, Sun Z N, et al. Wasserstein CNN: Learning invariant features  for  NIR-VIS  face  recognition. IEEE Trans Pattern Anal Mach Intell, 2019, 41（7）: 1761 [17] Shelhamer E, Long J, Darrell T. Fully convolutional networks for semantic  segmentation. IEEE Trans Pattern Anal Mach Intell, 2017, 39（4）: 640 [18] [19] LeCun  Y,  Bottou  L,  Bengio  Y,  et  al.  Gradient-based  learning applied to document recognition. Proc IEEE, 1998, 86（11）: 2278 Huang  G  B,  Zhu  Q  Y,  Siew  C  K.  Extreme  learning  machine: Theory and applications. Neurocomputing, 2006, 70（1-3）: 489 [20] Huang  G  B,  Chen  L,  Siew  C  K.  Universal  approximation  using incremental  constructive  feedforward  networks  with  random hidden nodes. IEEE Trans Neural Netw, 2006, 17（4）: 879 [21] Anthimopoulos M, Christodoulidis S, Ebner L, et al. Lung pattern classification  for  interstitial  lung  diseases  using  a  deep convolutional  neural  network. IEEE Trans Med Imaging,  2016, 35（5）: 1207 [22] Li J Y, Zhao Y K, Xue Z E, et al. A survey of model compression for deep neural networks. Chin J Eng, 2019, 41（10）: 1229 （李江昀, 赵义凯, 薛卓尔, 等. 深度神经网络模型压缩综述. 工 程科学学报, 2019, 41（10）：1229） [23] Ji S W, Xu W, Yang M, et al. 3D convolutional neural networks for  human  action  recognition. IEEE Trans Pattern Anal Mach Intell, 2013, 35（1）: 221 [24] Bartlett P L. The sample complexity of pattern classification with neural  networks:  The  size  of  the  weights  is  more  important  than the size of the network. IEEE Trans Inf Theory, 1998, 44（2）: 525 [25] PhysioNet.  MIT−BIH  Arrhythmias  Database  [J/OL]. PhysioNet Online (2020-04-15) [2021-01-12] https://physionet.org/files/mitdb/ 1.0.0 [26] Moody  G  B,  Mark  R  G.  The  impact  of  the  mit-bih  arrhythmia database. IEEE Eng Med Biol Mag, 2001, 20（3）: 45 [27] · 1232 · 工程科学学报，第 43 卷，第 9 期