期刊文献+

UGES反向传导算法:一种新的小样本深度机器学习模型 被引量:1

UGES backpropagation: a new deep machine learning model for small sample learning
下载PDF
导出
摘要 针对传统深度学习算法在样本不足时易出现过拟合的问题,提出了一类新的小样本深度学习模型:UGES反向传导模型。其基本思路是:在保留深层结构的同时,压缩需要学习参数的数量。作为一种与误差反向传导算法相容的间接编码模型,该算法对权值的随机分布特性进行重新编码,打破了不同隐含层之间的隔阂,并使用变分贝叶斯学习对网络进行全局训练。新模型的参数数目不再与输入变量维数及网络结构大小相关,同时强迫权值对于一定程度的扰动具有鲁棒性。最后,将所提出的算法用于外包软件项目风险识别这一典型的多维小样本问题中。对比实验表明,该模型达到了93.3%的样本外准确率,不仅保留了深度模型非线性表达能力,亦具备了小样本下优秀的泛化能力。 Original deep neural networks are prone to over-fit when the size of training examples To address this problem, a new efficient deep machine learning algorithm is proposed for small learning. The basic idea is to compress the number of parameters tO be learned while retaining structure. To achieve this, a back-prop compatible encoding scheme (i. e. UGES Back-prop is small. sample the deep ) is pro-
出处 《黑龙江大学自然科学学报》 CAS 北大核心 2015年第6期831-840,共10页 Journal of Natural Science of Heilongjiang University
基金 国家自然科学基金资助项目(71271061 70801020) 广东省大学生科技创新培育专项资金重点资助项目(308-GK151011) 广东省自然科学基金资助项目(2014A030313575) 广东省哲学社会科学"十二五"规划项目(GD12XGL14) 广东省公益研究与能力建设之软科学项目(2015A070704051) 广东省教育厅科技创新基金资助项目(2013KJCX0072) 广州市哲学社会科学发展"十二五"规划共建课题项目(14G41) 2014年国家级大学生创新训练计划项目(201411846001) 广东外语外贸大学教学研究重点A类项目(GYJYZDA14002)
关键词 深度机器学习 不确定性间接编码 变分贝叶斯学习 小样本问题 外包软件项目风险识别 deep machine learning uncertainty generative encoding varational bayesian learning small sample learning problems risk prediction of outsourcing software projects
  • 相关文献

参考文献49

  • 1BELLMAN R. Dynamic programming [ M]. Princeton, NJ: Princeton University Press, 1957. 被引量:1
  • 2HINTON G E, OSINDERO S, TEH Y W. A fast learning algorithm for deep belief nets[J]. Neural Computation, 2006, 18(7) : 1527 -1554. 被引量:1
  • 3VINCENT P, LAROCHELLE H, BENG10 Y, et al. Extracting and composing robust features with denoising autoencoders[ C ]. Proceedings of the 25th International Conference on Machine Learning. Helsinki: ACM, 2008:1096 -1103. 被引量:1
  • 4LEE H, GROSSE R, RANGANATH R, et al. Convolutional deep belief networks for scalable unsupervised learning of hierarchical representations [ C]. Proceedings of the 26th Annual International Conference on Machine Learning. Montreal: ACM, 2009:609 -616. 被引量:1
  • 5VINCENT P, LAROCHELLE H, LAJOIE I, et al. Stacked denoising autoencoders: learning useful representations in a deep network with a local denoising criterion [ J ]. The Journal of Machine Learning Research, 2010, 11 : 3371 - 3408. 被引量:1
  • 6EKANADHAM C. Sparse deep belief net models for visual area V2 [ J]. Advances in Neural Information Processing Systems, 2008:873 -880. 被引量:1
  • 7DENG J, ZHANG Z, MARCHI E, et al. Sparse autoeneoder-based feature transfer learning for speech emotion recognition [ C ]. Affective Compu- ting and Intelligent Interaction ( ACII), 2013 Humaine Association Conference on. Genova: IEEE, 2013 : 511 - 516. 被引量:1
  • 8HINTON G E, SRIVASTAVA N, KRIZHEVSKY A, et al. Improving neural networks by preventing co-adaptation of feature detectors[ J]. Impro- ving neural networks by preventing co-adaptation of feature detectors-ResearchGate, 2012, 3 (4) : 212 -223. 被引量:1
  • 9BLEI D M, JORDAN M I, PAISLEY J W. Variational bayesian inference with stochastic search [ J ]. Neuroimage, 2012,19 (3) : 727 -741. 被引量:1
  • 10KINGMA D P, WELLING M. Auto-encoding variational bayes [ J]. arXiv Preprint, arXiv: 1312. 6114, 2013. 被引量:1

同被引文献11

  • 1Ngiam J,Coates A,Lahiri A,et al. On optimization methods for deep learning[C]/ /Proceedings of the 28thInternational Con-ference on Machine Learning ( ICML-11) ,2011: 265-272. 被引量:1
  • 2Lee H,Pham P T,Largman Y,et al. Unsupervised feature learning for audio classification using convolutional deep belief net-works[C]/ / NIPS,2009: 1096-1104. 被引量:1
  • 3Kung S Y,Mak M W,Lin S H. Biometric Authentication: A Machine Learning Approach[M]. Beijing: Pearson Education,2005: 45-72. 被引量:1
  • 4Liu C,Wechsler H. Robust coding schemes for indexing and retrieval from large face databases[J]. IEEE Transactions onImage Processing,2000,9( 1) : 132-137. 被引量:1
  • 5Tolba A S,El-Baz A H,El-Harby A A. Face recognition: a literature review[J]. International Journal of Signal Processing,2005,2( 2) : 88-103. 被引量:1
  • 6Taigmany Y,Yang M,Ranzato M A,et al. Deepface: closing the gap to human-level performance in face verification [C]/ /Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition,2014: 1701-1708. 被引量:1
  • 7Hastie T,Tibshirani R. Classification by pairwise coupling[J]. Conference on Advances in Neural Information Processing Sys-tems,1998( 10) : 507-513. 被引量:1
  • 8Uglov J,Jakaite L,Schetinin V,et al. Comparing robustness of pairwise and multiclass neural-network systems for face recog-nition[J]. Hindawi Publishing Corporation EURASIP Journal on Advances in Signal Processing,2008( 7) : 1-7. 被引量:1
  • 9Schetinin V,Schult J,Scheidt B,et al. Learning multiclass neural-network models from electroencephalograms[J]. Knowl-edge-Based Intelligent Information and Engineering Systems,2003,2773: 155-162. 被引量:1
  • 10孙志军,薛磊,许阳明,王正.深度学习研究综述[J].计算机应用研究,2012,29(8):2806-2810. 被引量:628

引证文献1

二级引证文献2

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部