Can Pretext-Based Self-Supervised Learning Be Boosted by Downstream Data? A Theoretical Analysis

Jiaye Teng, Weiran Huang, Haowei He
Proceedings of The 25th International Conference on Artificial Intelligence and Statistics, PMLR 151:4198-4216, 2022.

Abstract

Pretext-based self-supervised learning learns the semantic representation via a handcrafted pretext task over unlabeled data and then uses the learned representation for downstream tasks, which effectively reduces the sample complexity of downstream tasks under Conditional Independence (CI) condition. However, the downstream sample complexity gets much worse if the CI condition does not hold. One interesting question is whether we can make the CI condition hold by using downstream data to refine the unlabeled data to boost self-supervised learning. At first glance, one might think that seeing downstream data in advance would always boost the downstream performance. However, we show that it is not intuitively true and point out that in some cases, it hurts the final performance instead. In particular, we prove both model-free and model-dependent lower bounds of the number of downstream samples used for data refinement. Moreover, we conduct various experiments on both synthetic and real-world datasets to verify our theoretical results.

Cite this Paper

BibTeX
@InProceedings{pmlr-v151-teng22a, title = { Can Pretext-Based Self-Supervised Learning Be Boosted by Downstream Data? A Theoretical Analysis }, author = {Teng, Jiaye and Huang, Weiran and He, Haowei}, booktitle = {Proceedings of The 25th International Conference on Artificial Intelligence and Statistics}, pages = {4198--4216}, 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/teng22a/teng22a.pdf}, url = {https://proceedings.mlr.press/v151/teng22a.html}, abstract = { Pretext-based self-supervised learning learns the semantic representation via a handcrafted pretext task over unlabeled data and then uses the learned representation for downstream tasks, which effectively reduces the sample complexity of downstream tasks under Conditional Independence (CI) condition. However, the downstream sample complexity gets much worse if the CI condition does not hold. One interesting question is whether we can make the CI condition hold by using downstream data to refine the unlabeled data to boost self-supervised learning. At first glance, one might think that seeing downstream data in advance would always boost the downstream performance. However, we show that it is not intuitively true and point out that in some cases, it hurts the final performance instead. In particular, we prove both model-free and model-dependent lower bounds of the number of downstream samples used for data refinement. Moreover, we conduct various experiments on both synthetic and real-world datasets to verify our theoretical results. } }
Endnote
%0 Conference Paper %T Can Pretext-Based Self-Supervised Learning Be Boosted by Downstream Data? A Theoretical Analysis %A Jiaye Teng %A Weiran Huang %A Haowei He %B Proceedings of The 25th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2022 %E Gustau Camps-Valls %E Francisco J. R. Ruiz %E Isabel Valera %F pmlr-v151-teng22a %I PMLR %P 4198--4216 %U https://proceedings.mlr.press/v151/teng22a.html %V 151 %X Pretext-based self-supervised learning learns the semantic representation via a handcrafted pretext task over unlabeled data and then uses the learned representation for downstream tasks, which effectively reduces the sample complexity of downstream tasks under Conditional Independence (CI) condition. However, the downstream sample complexity gets much worse if the CI condition does not hold. One interesting question is whether we can make the CI condition hold by using downstream data to refine the unlabeled data to boost self-supervised learning. At first glance, one might think that seeing downstream data in advance would always boost the downstream performance. However, we show that it is not intuitively true and point out that in some cases, it hurts the final performance instead. In particular, we prove both model-free and model-dependent lower bounds of the number of downstream samples used for data refinement. Moreover, we conduct various experiments on both synthetic and real-world datasets to verify our theoretical results.
APA
Teng, J., Huang, W. & He, H.. (2022). Can Pretext-Based Self-Supervised Learning Be Boosted by Downstream Data? A Theoretical Analysis . Proceedings of The 25th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 151:4198-4216 Available from https://proceedings.mlr.press/v151/teng22a.html.

Related Material