Papers of the day   All papers

Prototypical Contrastive Learning of Unsupervised Representations

Comments

Steven Hoi: Excited to introduce our Prototypical Contrastive Learning (PCL): unifies clustering and contrastive learning to push the frontiers of unsupervised learning. Blog: https://blog.einstein.ai/prototypical-contrastive-learning-pushing-the-frontiers-of-unsupervised-learning/ Paper: https://arxiv.org/abs/2005.04966 @SFResearch @LiJunnan0409 @CaimingXiong @RichardSocher https://t.co/ZCkqJIH3MQ

5 replies, 317 likes


Edward Dixon: #ContrastiveLearning, #SelfSupervisedLearning, now even more exciting! Come for the SOTA results, stay for the great explanations & hierarchical representations. From the nice people at @Salesforce: @lijunnan0409, @CaimingXiong, @RichardSocher, @stevenhoi https://arxiv.org/pdf/2005.04966.pdf https://t.co/5PQhVjbyQg

2 replies, 88 likes


Li Junnan: Code and pre-trained models for PCL are available! http://github.com/salesforce/PCL @stevenhoi @SFResearch @CaimingXiong @RichardSocher

0 replies, 23 likes


Daisuke Okanohara: For unsupervised representation learning, PCL performs k-means clustering on representations to obtain prototypes, then performs contrastive learning using the prototypes as targets. Set new SOTAs on unsupervised representation learning benchmarks. https://arxiv.org/abs/2005.04966

0 replies, 12 likes


arXiv in review: #NeurIPS2020 Prototypical Contrastive Learning of Unsupervised Representations. (arXiv:2005.04966v1 [cs\.CV]) http://arxiv.org/abs/2005.04966

0 replies, 7 likes


Steven Hoi: Unsupervised Learning is the holy grail of machine learning and AI. PCL brings together two schools of Unsupervised Learning: the classical school of Clustering and the emerging school of contrastive self-supervised learning. Established new SOTA results. Code will be released. https://t.co/BMeU2ussHJ

0 replies, 4 likes


Brundage Bot: Prototypical Contrastive Learning of Unsupervised Representations. Junnan Li, Pan Zhou, Caiming Xiong, Richard Socher, and Steven C. H. Hoi http://arxiv.org/abs/2005.04966

1 replies, 0 likes


Content

Found on May 16 2020 at https://arxiv.org/pdf/2005.04966.pdf

PDF content of a computer science paper: Prototypical Contrastive Learning of Unsupervised Representations