Semi-supervised Sequential Generative Models

被引:0
|
作者
Teng, Michael [1 ]
Le, Tuan Anh [2 ]
Scibior, Adam [3 ]
Wood, Frank [3 ,4 ]
机构
[1] Univ Oxford, Dept Engn Sci, Oxford, England
[2] MIT, Dept Brain & Cognit Sci, Cambridge, MA 02139 USA
[3] Univ British Columbia, Dept Comp Sci, Vancouver, BC, Canada
[4] Montreal Inst Learning Algorithms MILA, Montreal, PQ, Canada
基金
加拿大自然科学与工程研究理事会;
关键词
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
We introduce a novel objective for training deep generative time-series models with discrete latent variables for which supervision is only sparsely available. This instance of semi-supervised learning is challenging for existing methods, because the exponential number of possible discrete latent configurations results in high variance gradient estimators. We first overcome this problem by extending the standard semi-supervised generative modeling objective with reweighted wake-sleep. However, we find that this approach still suffers when the frequency of available labels varies between training sequences. Finally, we introduce a unified objective inspired by teacher-forcing and show that this approach is robust to variable length supervision. We call the resulting method caffeinated wake-sleep (CWS) to emphasize its additional dependence on real data. We demonstrate its effectiveness with experiments on MNIST, handwriting, and fruit fly trajectory data.
引用
收藏
页码:649 / 658
页数:10
相关论文
共 50 条
  • [1] Semi-supervised Learning with Deep Generative Models
    Kingma, Diederik P.
    Rezende, Danilo J.
    Mohamed, Shakir
    Welling, Max
    [J]. ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 27 (NIPS 2014), 2014, 27
  • [2] A Semi-supervised AUC Optimization Method with Generative Models
    Fujino, Akinori
    Ueda, Naonori
    [J]. 2016 IEEE 16TH INTERNATIONAL CONFERENCE ON DATA MINING (ICDM), 2016, : 883 - 888
  • [3] Semi-Supervised Analysis of the Electrocardiogram Using Deep Generative Models
    Rasmussen, Soren M.
    Jensen, Malte E. K.
    Meyhoff, Christian S.
    Aasvang, Eske K.
    Sorensen, Helge B. D.
    [J]. 2021 43RD ANNUAL INTERNATIONAL CONFERENCE OF THE IEEE ENGINEERING IN MEDICINE & BIOLOGY SOCIETY (EMBC), 2021, : 1124 - 1127
  • [4] Learning Disentangled Representations with Semi-Supervised Deep Generative Models
    Siddharth, N.
    Paige, Brooks
    van de Meent, Jan-Willem
    Desmaison, Alban
    Goodman, Noah D.
    Kohli, Pushmeet
    Wood, Frank
    Torr, Philip H. S.
    [J]. ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 30 (NIPS 2017), 2017, 30
  • [5] Semi-Supervised Generative Models for Multi-Agent Trajectories
    Fassmeyer, Dennis
    Fassmeyer, Pascal
    Brefeld, Ulf
    [J]. ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 35 (NEURIPS 2022), 2022,
  • [6] Optimization of semi-supervised generative adversarial network models: a survey
    Ma, Yongqing
    Zheng, Yifeng
    Zhang, Wenjie
    Wei, Baoya
    Lin, Ziqiong
    Liu, Weiqiang
    Li, Zhehan
    [J]. INTERNATIONAL JOURNAL OF INTELLIGENT COMPUTING AND CYBERNETICS, 2024,
  • [7] Unbiased Generative Semi-Supervised Learning
    Fox-Roberts, Patrick
    Rosten, Edward
    [J]. JOURNAL OF MACHINE LEARNING RESEARCH, 2014, 15 : 367 - 443
  • [8] A Study of Semi-supervised Generative Ensembles
    Zanda, Manuela
    Brown, Gavin
    [J]. MULTIPLE CLASSIFIER SYSTEMS, PROCEEDINGS, 2009, 5519 : 242 - 251
  • [9] Generative Adversarial Training for Supervised and Semi-supervised Learning
    Wang, Xianmin
    Li, Jing
    Liu, Qi
    Zhao, Wenpeng
    Li, Zuoyong
    Wang, Wenhao
    [J]. FRONTIERS IN NEUROROBOTICS, 2021, 15
  • [10] Semi-Supervised Sequential Kernel Regression Models with Penalty Functions
    Tang, Hengjin
    Miyamoto, Sadaaki
    Endo, Yasunori
    [J]. JOURNAL OF ADVANCED COMPUTATIONAL INTELLIGENCE AND INTELLIGENT INFORMATICS, 2015, 19 (01) : 49 - 55