Page 48 - 201901

44 2019 年 1 月

[5] SreeB R L, Vijaya M S. Building acoustic model for [11] Cho K, Raiko T, Llin A. Parallel tempering is efficient for
phoneme recognition using PSO-DBN[J]. International learning restricted Boltzmann machines[C]. Neural Net-
Journal of Business Intelligence & Data Mining, 2018, works (IJCNN), The 2010 International Joint Conference
1(1): 1. on, IEEE, 2010.
[6] Mohamed A R, Sainath T N, Dahl G, et al. Deep belief [12] Desjardins G, Courville A, Bengio Y. Adaptive parallel
networks using discriminative features for phone recog- tempering for stochastic maximum likelihood learning of
nition[C]. IEEE International Conference on Acoustics, RBMs[J]. Computer Science, 2010, arXiv: 1012.3476.
Speech and Signal Processing, 2011. [13] Koller D, Friedman N. Probabilistic graphical models:
[7] Hinton G E. A practical guide to training restricted Boltz- principles and techniques-adaptive computation and ma-
mann machines[J]. Momentum, 2012, 9(1): 599–619. chine learning[M]. Massachusetts: The MIT Press, 2009:
[8] Jang H, Choi H, Yi Y, et al. Adiabatic persistent con- 161–168.
trastive divergence learning[C]. IEEE International Sym- [14] He F, Han Y, Wang H, et al. Deep learning architec-
posium on Information Theory, 2017. ture for iris recognition based on optimal Gabor filters
[9] Hinton G E, Osindero S, Teh Y W. A fast learning algo- and deep belief network[J]. Journal of Electronic Imag-
rithm for deep belief nets[J]. Neural Computation, 2006, ing, 2017, 26(2): 023005.
18(7): 1527–1554. [15] Deng J, Xu X, Zhang Z, et al. Semi-supervised au-
[10] Berglund M, Raiko T. Stochastic gradient estimate vari- toencoders for speech emotion recognition[J]. IEEE/ACM
ance in contrastive divergence and persistent contrastive Transactions on Audio Speech & Language Processing,
divergence[J]. Computer Science, 2014, arXiv: 1312.6002. 2017, 26(1): 31–43.