English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

For an image query, unsupervised contrastive learning labels crops of thesame image as positives, and other image crops as negatives. Althoughintuitive, such a native label assignment strategy cannot reveal the underlyingsemantic similarity between a query and its positives and negatives, andimpairs performance, since some negatives are semantically similar to the queryor even share the same semantic class as the query. In this work, we firstprove that for contrastive learning, inaccurate label assignment heavilyimpairs its generalization for semantic instance discrimination, while accuratelabels benefit its generalization. Inspired by this theory, we propose a novelself-labeling refinement approach for contrastive learning. It improves thelabel quality via two complementary modules: (i) self-labeling refinery (SLR)to generate accurate labels and (ii) momentum mixup (MM) to enhance similaritybetween query and its positive. SLR uses a positive of a query to estimatesemantic similarity between a query and its positive and negatives, andcombines estimated similarity with vanilla label assignment in contrastivelearning to iteratively generate more accurate and informative soft labels. Wetheoretically show that our SLR can exactly recover the true semantic labels oflabel-corrupted data, and supervises networks to achieve zero prediction erroron classification tasks. MM randomly combines queries and positives to increasesemantic similarity between the generated virtual queries and their positivesso as to improves label accuracy. Experimental results on CIFAR10, ImageNet,VOC and COCO show the effectiveness of our method. PyTorch code and model willbe released online.

A Theory-Driven Self-Labeling Refinement Method for Contrastive Representation Learning