Surrogate- and invariance-boosted contrastive learning for data-scarce applications in science

Deep learning techniques usually require a large quantity of training data and may be challenging for scarce datasets. The authors propose a framework that involves contrastive and transfer learning and reduces data requirements for training while keeping the prediction accuracy.

Bibliographic Details
Main Authors: Charlotte Loh, Thomas Christensen, Rumen Dangovski, Samuel Kim, Marin Soljačić
Format: Article
Language:English
Published: Nature Portfolio 2022-07-01
Series:Nature Communications
Online Access:https://doi.org/10.1038/s41467-022-31915-y
_version_ 1818190148610818048
author Charlotte Loh
Thomas Christensen
Rumen Dangovski
Samuel Kim
Marin Soljačić
author_facet Charlotte Loh
Thomas Christensen
Rumen Dangovski
Samuel Kim
Marin Soljačić
author_sort Charlotte Loh
collection DOAJ
description Deep learning techniques usually require a large quantity of training data and may be challenging for scarce datasets. The authors propose a framework that involves contrastive and transfer learning and reduces data requirements for training while keeping the prediction accuracy.
first_indexed 2024-12-11T23:54:07Z
format Article
id doaj.art-07829645c7eb4f6cb3801a7892810430
institution Directory Open Access Journal
issn 2041-1723
language English
last_indexed 2024-12-11T23:54:07Z
publishDate 2022-07-01
publisher Nature Portfolio
record_format Article
series Nature Communications
spelling doaj.art-07829645c7eb4f6cb3801a78928104302022-12-22T00:45:24ZengNature PortfolioNature Communications2041-17232022-07-0113111210.1038/s41467-022-31915-ySurrogate- and invariance-boosted contrastive learning for data-scarce applications in scienceCharlotte Loh0Thomas Christensen1Rumen Dangovski2Samuel Kim3Marin Soljačić4Department of Electrical Engineering and Computer Science, Massachusetts Institute of TechnologyDepartment of Physics, Massachusetts Institute of TechnologyDepartment of Electrical Engineering and Computer Science, Massachusetts Institute of TechnologyDepartment of Electrical Engineering and Computer Science, Massachusetts Institute of TechnologyDepartment of Physics, Massachusetts Institute of TechnologyDeep learning techniques usually require a large quantity of training data and may be challenging for scarce datasets. The authors propose a framework that involves contrastive and transfer learning and reduces data requirements for training while keeping the prediction accuracy.https://doi.org/10.1038/s41467-022-31915-y
spellingShingle Charlotte Loh
Thomas Christensen
Rumen Dangovski
Samuel Kim
Marin Soljačić
Surrogate- and invariance-boosted contrastive learning for data-scarce applications in science
Nature Communications
title Surrogate- and invariance-boosted contrastive learning for data-scarce applications in science
title_full Surrogate- and invariance-boosted contrastive learning for data-scarce applications in science
title_fullStr Surrogate- and invariance-boosted contrastive learning for data-scarce applications in science
title_full_unstemmed Surrogate- and invariance-boosted contrastive learning for data-scarce applications in science
title_short Surrogate- and invariance-boosted contrastive learning for data-scarce applications in science
title_sort surrogate and invariance boosted contrastive learning for data scarce applications in science
url https://doi.org/10.1038/s41467-022-31915-y
work_keys_str_mv AT charlotteloh surrogateandinvarianceboostedcontrastivelearningfordatascarceapplicationsinscience
AT thomaschristensen surrogateandinvarianceboostedcontrastivelearningfordatascarceapplicationsinscience
AT rumendangovski surrogateandinvarianceboostedcontrastivelearningfordatascarceapplicationsinscience
AT samuelkim surrogateandinvarianceboostedcontrastivelearningfordatascarceapplicationsinscience
AT marinsoljacic surrogateandinvarianceboostedcontrastivelearningfordatascarceapplicationsinscience