We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.LG

Change to browse by:

cs

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo

Computer Science > Machine Learning

Title: Bayesian Self-Supervised Contrastive Learning

Abstract: Recent years have witnessed many successful applications of contrastive learning in diverse domains, yet its self-supervised version still remains many exciting challenges. As the negative samples are drawn from unlabeled datasets, a randomly selected sample may be actually a false negative to an anchor, leading to incorrect encoder training. This paper proposes a new self-supervised contrastive loss called the BCL loss that still uses random samples from the unlabeled data while correcting the resulting bias with importance weights. The key idea is to design the desired sampling distribution for sampling hard true negative samples under the Bayesian framework. The prominent advantage lies in that the desired sampling distribution is a parametric structure, with a location parameter for debiasing false negative and concentration parameter for mining hard negative, respectively. Experiments validate the effectiveness and superiority of the BCL loss.
Comments: 24 pages
Subjects: Machine Learning (cs.LG)
Cite as: arXiv:2301.11673 [cs.LG]
  (or arXiv:2301.11673v4 [cs.LG] for this version)

Submission history

From: Bin Liu [view email]
[v1] Fri, 27 Jan 2023 12:13:06 GMT (2093kb,D)
[v2] Thu, 2 Feb 2023 03:22:19 GMT (1692kb,D)
[v3] Tue, 23 May 2023 07:53:05 GMT (2394kb,D)
[v4] Wed, 31 Jan 2024 03:02:06 GMT (2480kb,D)

Link back to: arXiv, form interface, contact.