Contrastive Learning Inverts the Data Generating Process

Published in ICML 2021, 2020

Zimmermann, R. S., Sharma, Y., Schneider, S. Bethge, M. and Brendel, W., Contrastive Learning Inverts the Data Generating Process.

Contrastive learning has recently seen tremendous success in self-supervised learning. So far, however, it is largely unclear why the learned representations generalize so effectively to a large variety of downstream tasks. We here prove that feedforward models trained with objectives belonging to the commonly used InfoNCE family learn to implicitly invert the underlying generative model of the observed data. While the proofs make certain statistical assumptions about the generative model, we observe empirically that our findings hold even if these assumptions are severely violated. Our theory highlights a fundamental connection between contrastive learning, generative modeling, and nonlinear independent component analysis, thereby furthering our understanding of the learned representations as well as providing a theoretical foundation to derive more effective contrastive losses.

Project website    Full paper

@article{zimmermann2021contrastive, @article{zimmermann2021cl,
  author = {
    Zimmermann, Roland S. and
    Sharma, Yash and
    Schneider, Steffen and
    Bethge, Matthias and
    Brendel, Wieland
  },
  title = {
    Contrastive Learning Inverts
    the Data Generating Process
  },
booktitle = {Proceedings of the 38th International Conference on Machine Learning,
    {ICML} 2021, 18-24 July 2021, Virtual Event},
  series = {Proceedings of Machine Learning Research},
  volume = {139},
  pages = {12979--12990},
  publisher = {{PMLR}},
  year = {2021},
  url = {http://proceedings.mlr.press/v139/zimmermann21a.html},
}