2020

Weakly-Supervised Disentanglement Without Compromises

Locatello, Francesco, Poole, Ben, Rätsch, Gunnar et al.

Understand

Intelligent agents should be able to learn useful representations by observing changes in their environment.

We model such observations as pairs of non-i.i.d.
images sharing at least one of the underlying factors of variation.
First, we theoretically show that only knowing how many factors have changed, but not which ones, is sufficient to learn disentangled representations.

Built on

Causality for machine learning, 2019
Original
Schölkopf, B · 1911
Earlier work this paper cites.
Standardization of progressive matrices, 1938
Raven, J. C · 1941
Earlier work this paper cites.
Learning invariance from transformation sequences
Földiák, P · 1991
Earlier work this paper cites.
Improving generalization for temporal difference learning: The successor representation
Dayan, P · 1993
Earlier work this paper cites.
Independent component analysis, a new concept?
Comon, P · 1994
Earlier work this paper cites.
Reinforcement driven information acquisition in non-deterministic environments
Storck, J., Hochreiter, S., and Schmidhuber, J · 1995
Earlier work this paper cites.
The application of cluster analysis in strategic management research: an analysis and critique
Ketchen, D. J. and Shook, C. L · 1996
Earlier work this paper cites.
Feature extraction through lococode
Hochreiter, S. and Schmidhuber, J · 1999
Earlier work this paper cites.
Nonlinear independent component analysis: Existence and uniqueness results
Hyvärinen, A. and Pajunen, P · 1999
Earlier work this paper cites.
Improving predictive inference under covariate shift by weighting the log-likelihood function
Shimodaira, H · 2000
Earlier work this paper cites.
Learning methods for generic object recognition with invariance to pose and lighting
LeCun, Y., Huang, F. J., and Bottou, L · 2004
Earlier work this paper cites.
Learning graphical model structure using l1-regularization paths
Schmidt, M., Niculescu-Mizil, A., Murphy, K., et al · 2007
Earlier work this paper cites.
Building classifiers with independency constraints
Calders, T., Kamiran, F., and Pechenizkiy, M · 2009
Earlier work this paper cites.
Dataset shift in machine learning
Quionero-Candela, J., Sugiyama, M., Schwaighofer, A., and Lawrence, N. D · 2009
Earlier work this paper cites.
Impossibility theorems for domain adaptation
Ben-David, S., Lu, T., Luu, T., and Pál, D · 2010
Earlier work this paper cites.
On causal and anticausal learning
Schölkopf, B., Janzing, D., Peters, J., Sgouritsa, E., Zhang, K., and Mooij, J · 2012
Earlier work this paper cites.
Representation learning: A review and new perspectives
Bengio, Y., Courville, A., and Vincent, P · 2013
Earlier work this paper cites.
Discovering hidden factors of variation in deep networks
Original
Cheung, B., Livezey, J. A., Bansal, A. K., and Olshausen, B. A · 2014
Earlier work this paper cites.
Learning to disentangle factors of variation with manifold interaction
Reed, S., Sohn, K., Zhang, Y., and Lee, H · 2014
Earlier work this paper cites.
Learning to linearize under uncertainty
Goroshin, R., Mathieu, M. F., and LeCun, Y · 2015
Earlier work this paper cites.
Bayesian representation learning with oracle constraints
Original
Karaletsos, T., Belongie, S., and Rätsch, G · 2015
Earlier work this paper cites.
Deep convolutional inverse graphics network
Kulkarni, T. D., Whitney, W. F., Kohli, P., and Tenenbaum, J · 2015
Earlier work this paper cites.
Deep visual analogy-making
Reed, S., Zhang, Y., Zhang, Y., and Lee, H · 2015
Earlier work this paper cites.
Weakly-supervised disentangling with recurrent transformations for 3D view synthesis
Yang, J., Reed, S. E., Yang, M.-H., and Lee, H · 2015
Earlier work this paper cites.
Multi-source domain adaptation: A causal view
Zhang, K., Gong, M., and Schölkopf, B · 2015
Earlier work this paper cites.

Similar

On the relation between accuracy and fairness in binary classification
Original
Zliobaite, I · 2015
Cited alongside, same era.
Unsupervised feature extraction by time-contrastive learning and nonlinear ica
Hyvarinen, A. and Morioka, H · 2016
Cited alongside, same era.
Disentangling factors of variation in deep representation using adversarial training
Mathieu, M. F., Zhao, J. J., Ramesh, A., Sprechmann, P., and LeCun, Y · 2016
Cited alongside, same era.
A survey of inductive biases for factorial representation-learning
Original
Ridgeway, K · 2016
Cited alongside, same era.
Understanding visual concepts with continuation learning
Original
Whitney, W. F., Chang, M., Kulkarni, T., and Tenenbaum, J. B · 2016
Cited alongside, same era.
The consciousness prior
Original
Bengio, Y · 2017
Cited alongside, same era.
Factorized variational autoencoders for modeling audience reactions to movies
Deng, Z., Navarathna, R., Carr, P., Mandt, S., Yue, Y., Matthews, I., and Mori, G · 2017
Cited alongside, same era.
Unsupervised learning of disentangled representations from video
Denton, E. L. and Birodkar, V · 2017
Cited alongside, same era.
A disentangled recognition and nonlinear dynamics model for unsupervised learning
Fraccaro, M., Kamronn, S., Paquet, U., and Winther, O · 2017
Cited alongside, same era.
Unsupervised learning of disentangled and interpretable representations from sequential data
Hsu, W.-N., Zhang, Y., and Glass, J · 2017
Cited alongside, same era.
Behind distribution shift: Mining driving forces of changes and causal arrows
Huang, B., Zhang, K., Zhang, J., Sanchez-Romero, R., Glymour, C., and Schölkopf, B · 2017
Cited alongside, same era.
Nonlinear ica of temporally dependent stationary sources
Hyvarinen, A. and Morioka, H · 2017
Cited alongside, same era.

Then

Variational inference of disentangled latent concepts from unlabeled observations
Kumar, A., Sattigeri, P., and Balakrishnan, A · 2018
Later among the works it cites.
Competitive training of mixtures of independent deep generative models
Locatello, F., Vincent, D., Tolstikhin, I., Rätsch, G., Gelly, S., and Schölkopf, B · 2018
Later among the works it cites.
Learning deep disentangled embeddings with the f-statistic loss
Ridgeway, K. and Mozer, M. C · 2018
Later among the works it cites.
Measuring abstract reasoning in neural networks
Santoro, A., Hill, F., Barrett, D., Morcos, A., and Lillicrap, T · 2018
Later among the works it cites.
Recent advances in autoencoder-based representation learning
Original
Tschannen, M., Bachem, O., and Lucic, M · 2018
Later among the works it cites.
Disentangled sequential autoencoder
Yingzhen, L. and Mandt, S · 2018
Later among the works it cites.
A meta-transfer objective for learning to disentangle causal mechanisms
Original
Bengio, Y., Deleu, T., Rahaman, N., Ke, R., Lachapelle, S., Bilaniuk, O., Goyal, A., and Pal, C · 2019
Later among the works it cites.
Hybrid deep fault detection and isolation: Combining deep neural networks and system performance models
Original
Chao, M. A., Kulkarni, C., Goebel, K., and Fink, O · 2019
Later among the works it cites.
Disentangled representation learning in cardiac image analysis
Chartsias, A., Joyce, T., Papanastasiou, G., Semple, S., Williams, M., Newby, D. E., Dharmakumar, R., and Tsaftaris, S. A · 2019
Later among the works it cites.
Flexibly fair representation learning by disentanglement
Creager, E., Madras, D., Jacobsen, J.-H., Weis, M., Swersky, K., Pitassi, T., and Zemel, R · 2019
Later among the works it cites.
A heuristic for unsupervised model selection for variational disentangled representation learning
Original
Duan, S., Watters, N., Matthey, L., Burgess, C. P., Lerchner, A., and Higgins, I · 2019
Later among the works it cites.
Deep self-organization: Interpretable discrete representation learning on time series
Fortuin, V., Hüser, M., Locatello, F., Strathmann, H., and Rätsch, G · 2019
Later among the works it cites.
On the transfer of inductive bias from simulation to the real world: a new disentanglement dataset
Gondal, M. W., Wüthrich, M., Miladinović, D., Locatello, F., Breidt, M., Volchkov, V., Akpo, J., Bachem, O., Schölkopf, B., and Bauer, S · 2019
Later among the works it cites.
The incomplete rosetta stone problem: Identifiability results for multi-view nonlinear ica
Gresele, L., Rubenstein, P. K., Mehrjou, A., Locatello, F., and Schölkopf, B · 2019
Later among the works it cites.
Group-based learning of disentangled representations with generalizability for novel contents
Hosoya, H · 2019
Later among the works it cites.
Nonlinear ica using auxiliary variables and generalized contrastive learning
Hyvarinen, A., Sasaki, H., and Turner, R. E · 2019
Later among the works it cites.
Variational autoencoders and nonlinear ICA: A unifying framework
Original
Khemakhem, I., Kingma, D. P., and Hyvärinen, A · 2019
Later among the works it cites.
Interventional robustness of deep latent variable models
Suter, R., Miladinović, D., Bauer, S., and Schölkopf, B · 2019
Later among the works it cites.
Are disentangled representations helpful for abstract visual reasoning?
van Steenkiste, S., Locatello, F., Schmidhuber, J., and Bachem, O · 2019
Later among the works it cites.
Weakly supervised disentanglement by pairwise similarities
Chen, J. and Batmanghelich, K · 2020
Closest in time.
Discovering physical concepts with neural networks
Iten, R., Metger, T., Wilming, H., Del Rio, L., and Renner, R · 2020
Closest in time.
Disentangling factors of variation using few labels
Locatello, F., Tschannen, M., Bauer, S., Rätsch, G., Schölkopf, B., and Bachem, O · 2020
Closest in time.
Weakly supervised disentanglement with guarantees
Shu, R., Chen, Y., Kumar, A., Ermon, S., and Poole, B · 2020
Closest in time.
Disentanglement by nonlinear ICA with general incompressible-flow networks (GIN)
Original
Sorrenson, P., Rother, C., and Köthe, U · 2020
Closest in time.

Beyond the bibliography

alphaXiv searches the wider corpus for related work and actual follow-ups.

Open on alphaXiv

alphaXiv is searching for related work…

Reading the bibliography…