2KR1ZHEVSKY A, SUTSKEVER I, HINTON G E. hnagenet classification with decd convolutional neura nelworks[C]//Advances in Neural Infornmtion Pro cessing Syslems. Red Hook. NY: Curran Associates. 2012:1097-1105.
3DAHI. G E,YU D.DENG L,et al. Contexl-dependent pre-lrained deep neural networks for large-wmabulary speech rccognition[J]. Audio. Speech. and l.anguage Processing. IEEE Transactions on, 2012.20( 1 ) :30- 12.
4ER H A N D. COURVII.I.E A. BENGIC Y. Understand ing representations learned in deep architectures,TR1355[R/OL. (2010-10-19)[2016-06-11]. http://www. dumitru, ca/files/publicalions/invarianes_techre- porl. pdf.
5HINTON G E.SALAKHUTI)INOV R. Reducing thedimensionality of data with neural networks [J]. Sci- ence, 2006,313(5786) :504-507.
6SOOMRO K,ZAM1R A R,SHAN M. Ucfl01a data- set of 101 human actions classes from videos in the wild[J].CoRR, 2012 : abs/1212. 0402.
7DANL G E, SAINATH T N, HINTON G E. Improving deep neural networks for LVCSR Using rectified linear u- nits and dropout[C]//Acoustics, Speech and Signal Pro- cessing (ICASSP), 2013 IEEE International Conference on. Piscataway, NJ : IEEE, 2013 : 8609-8613.
8LE Q V, RAMZATO M, MONGA R, et al. Building high-level features using large scale unsupervised learning,Ill2. 6209[R]. New York, USA: Cornell U- niversity, 2012.
9HINTON G E, OSINDERO S, TEH Y W. A fast learning algorithm for deep belief nets[J]. Neural Com- putation,2006,18(7) :1527-1554.
10GLOROT X, BORDES A, BENGTO Y. Deep sparse ) rectifier networks[C]. JMLR Workshoop and Confer- ence Proceedings : AISTATS 2011. Brookline, MA: Microtome Publishing, 2011 : 315-323.
