Independent Subspace Analysis for Unsupervised Learning of Disentangled Representations

被引:0
|
作者
Stuhmer, Jan [1 ]
Turner, Richard E. [2 ]
Nowozin, Sebastian [3 ]
机构
[1] Microsoft Res, Redmond, WA 98052 USA
[2] Univ Cambridge, Cambridge, England
[3] Microsoft Res, Google Brain, Redmond, WA USA
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
Recently there has been an increased interest in unsupervised learning of disentangled representations using the Variational Autoencoder (VAE) framework. Most of the existing work has focused largely on modifying the variational cost function to achieve this goal. We first show that these modifications, e.g. beta-VAE, simplify the tendency of variational inference to underfit causing pathological over-pruning and over-orthogonalization of learned components. Second we propose a complementary approach: to modify the probabilistic model with a structured latent prior. This prior allows to discover latent variable representations that are structured into a hierarchy of independent vector spaces. The proposed prior has three major advantages: First, in contrast to the standard VAE normal prior the proposed prior is not rotationally invariant. This resolves the problem of unidentifiability of the standard VAE normal prior. Second, we demonstrate that the proposed prior encourages a disentangled latent representation which facilitates learning of disentangled representations. Third, extensive quantitative experiments demonstrate that the prior significantly mitigates the trade-off between reconstruction loss and disentanglement over the state of the art.
引用
收藏
页数:10
相关论文
共 50 条
  • [1] A Commentary on the Unsupervised Learning of Disentangled Representations
    Locatello, Francesco
    Bauer, Stefan
    Lucie, Mario
    Raetsch, Gunnar
    Gelly, Sylvain
    Schoelkopf, Bernhard
    Bachem, Olivier
    [J]. THIRTY-FOURTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, THE THIRTY-SECOND INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE AND THE TENTH AAAI SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE, 2020, 34 : 13681 - 13684
  • [2] Unsupervised Learning of Disentangled Representations from Video
    Denton, Emily
    Birodkar, Vighnesh
    [J]. ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 30 (NIPS 2017), 2017, 30
  • [3] A Sober Look at the Unsupervised Learning of Disentangled Representations and their Evaluation
    Locatello, Francesco
    Bauer, Stefan
    Lucic, Mario
    Ratsch, Gunnar
    Gelly, Sylvain
    Schoelkopf, Bernhard
    Bachem, Olivier
    [J]. JOURNAL OF MACHINE LEARNING RESEARCH, 2020, 21
  • [4] Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representations
    Locatello, Francesco
    Bauer, Stefan
    Lucic, Mario
    Ratsch, Gunnar
    Gelly, Sylvain
    Scholkopf, Bernhard
    Bachem, Olivier
    [J]. INTERNATIONAL CONFERENCE ON MACHINE LEARNING, VOL 97, 2019, 97
  • [5] Learning Disentangled Representations via Independent Subspaces
    Awiszus, Maren
    Ackermann, Hanno
    Rosenhahn, Bodo
    [J]. 2019 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION WORKSHOPS (ICCVW), 2019, : 560 - 568
  • [6] Geometric Inductive Biases for Identifiable Unsupervised Learning of Disentangled Representations
    Pan, Ziqi
    Niu, Li
    Zhang, Liqing
    [J]. THIRTY-SEVENTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, VOL 37 NO 8, 2023, : 9372 - 9380
  • [7] Unsupervised Learning of Disentangled and Interpretable Representations from Sequential Data
    Hsu, Wei-Ning
    Zhang, Yu
    Glass, James
    [J]. ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 30 (NIPS 2017), 2017, 30
  • [8] Linear Disentangled Representations and Unsupervised Action Estimation
    Painter, Matthew
    Hare, Jonathon
    Prugel-Bennett, Adam
    [J]. ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 33, NEURIPS 2020, 2020, 33
  • [9] Unsupervised MR harmonization by learning disentangled representations using information bottleneck theory
    Zuo, Lianrui
    Dewey, Blake E.
    Liu, Yihao
    He, Yufan
    Newsome, Scott D.
    Mowry, Ellen M.
    Resnick, Susan M.
    Prince, Jerry L.
    Carass, Aaron
    [J]. NEUROIMAGE, 2021, 243
  • [10] Unsupervised learning of disentangled representations in deep restricted kernel machines with orthogonality constraints
    Tonin, Francesco
    Patrinos, Panagiotis
    Suykens, Johan A. K.
    [J]. NEURAL NETWORKS, 2021, 142 : 661 - 679