AISTATS 2022poster36 citations

Contrasting the landscape of contrastive and non-contrastive learning

Ashwini Pokle, Jinjin Tian, Yuchen Li, Andrej Risteski

Abstract

A lot of recent advances in unsupervised feature learning are based on designing features which are invariant under semantic data augmentations. A common way to do this is contrastive learning, which uses positive and negative samples. Some recent works however have shown promising results for non-contrastive learning, which does not require negative samples. However, the non-contrastive losses have obvious “collapsed” minima, in which the encoders output a constant feature embedding, independent of the input. A folk conjecture is that so long as these collapsed solutions are avoided, the produced feature representations should be good. In our paper, we cast doubt on this story: we show through theoretical results and controlled experiments that even on simple data models, non-contrastive losses have a preponderance of non-collapsed bad minima. Moreover, we show that the training process does not avoid these minima. Code for this work can be found at https://github.com/ashwinipokle/contrastive_landscape.

BibTeX
@InProceedings{pmlr-v151-pokle22a,
  title = 	 { Contrasting the landscape of contrastive and non-contrastive learning },
  author =       {Pokle, Ashwini and Tian, Jinjin and Li, Yuchen and Risteski, Andrej},
  booktitle = 	 {Proceedings of The 25th International Conference on Artificial Intelligence and Statistics},
  pages = 	 {8592--8618},
  year = 	 {2022},
  editor = 	 {Camps-Valls, Gustau and Ruiz, Francisco J. R. and Valera, Isabel},
  volume = 	 {151},
  series = 	 {Proceedings of Machine Learning Research},
  month = 	 {28--30 Mar},
  publisher =    {PMLR},
  pdf = 	 {https://proceedings.mlr.press/v151/pokle22a/pokle22a.pdf},
  url = 	 {https://proceedings.mlr.press/v151/pokle22a.html},
  abstract = 	 { A lot of recent advances in unsupervised feature learning are based on designing features which are invariant under semantic data augmentations. A common way to do this is contrastive learning, which uses positive and negative samples. Some recent works however have shown promising results for non-contrastive learning, which does not require negative samples. However, the non-contrastive losses have obvious “collapsed” minima, in which the encoders output a constant feature embedding, independent of the input. A folk conjecture is that so long as these collapsed solutions are avoided, the produced feature representations should be good. In our paper, we cast doubt on this story: we show through theoretical results and controlled experiments that even on simple data models, non-contrastive losses have a preponderance of non-collapsed bad minima. Moreover, we show that the training process does not avoid these minima. Code for this work can be found at https://github.com/ashwinipokle/contrastive_landscape. }
}
Contrasting the landscape of contrastive and non-contrastive learning · AISTATS 2022