Papers of the day   All papers

Supervised Contrastive Learning

Comments

Dilip Krishnan: New paper on *Supervised Contrastive Learning*: https://arxiv.org/abs/2004.11362 A new loss function to train supervised deep networks, based on contrastive learning! Our new loss performs significantly better than cross-entropy across a range of architectures and data augmentations. https://t.co/CTnbf1ruwT

9 replies, 381 likes


Dilip Krishnan: Reference PyTorch code released by @YonglongT for our paper on Supervised Contrastive Learning (https://arxiv.org/abs/2004.11362)! Find it here: https://github.com/HobbitLong/SupContrast https://t.co/iL4Lhmy3T5

4 replies, 269 likes


Phillip Isola: Even wonder if all the fancy new contrastive objectives would be useful for regular old *supervised* learning? Turns out they can be!

4 replies, 118 likes


roadrunner01: Supervised Contrastive Learning pdf: https://arxiv.org/pdf/2004.11362.pdf abs: https://arxiv.org/abs/2004.11362 https://t.co/D3rkUqwVtu

0 replies, 71 likes


mat kelcey: Haha. We did exactly this with some client work last year. I thought it was just a known approach ?

3 replies, 49 likes


Daisuke Okanohara: Contrastive loss used in self-supervised learning is also effective for supervised learning (extended for supporting multiple positive examples), improving the accuracy, the robustness for corruption, and the stability against hype parameter changes. https://arxiv.org/abs/2004.11362

0 replies, 11 likes


David Praise Chukwuma Kalu: Supervised Contrastive Learning #DavidPraiseKalu #artificialintelligence #BigData #DataScience #AI #MachineLearning #DeepLearning #Python #CloudComputing #DataScientist #Linux #Programming #Coding #100DaysofCode #SupervisedLearning #ContrastiveLearning https://arxiv.org/abs/2004.11362 https://t.co/fJBz11XZdn

0 replies, 9 likes


Dilip Krishnan: It shows clear benefits in top-1 accuracy and robustness; and is more stable across a range of hyperparameters. Joint work with @PrannayKhosla @YonglongT @phillip_isola and other colleagues. https://t.co/AU4w68k9OA

0 replies, 8 likes


David Praise Chukwuma Kalu: Supervised Contrastive Learning #DavidPraiseKalu #artificialintelligence #BigData #DataScience #AI #MachineLearning #DeepLearning #Python #CloudComputing #DataScientist #Linux #Programming #Coding #100DaysofCode #SupervisedLearning #ContrastiveLearning https://arxiv.org/abs/2004.11362 https://t.co/AJzjblGb5w

0 replies, 7 likes


Prannay Khosla: New paper out: https://arxiv.org/abs/2004.11362 Supervised Contrastive Learning TLDR; representation learning by leveraging labels leads to more robust models and boosts in classification accuracy.

1 replies, 7 likes


Yonglong Tian: what will happen if we build views that only share label info in the contrastive learning framework? Check out this new work... https://arxiv.org/pdf/2004.11362.pdf

0 replies, 7 likes


MONTREAL.AI: Supervised Contrastive Learning Khosla et al.: https://arxiv.org/abs/2004.11362 #SupervisedLearning #MachineLearning #ContrastiveLearning https://t.co/ytv9S27qgn

0 replies, 3 likes


Content

Found on Apr 24 2020 at https://arxiv.org/pdf/2004.11362.pdf

PDF content of a computer science paper: Supervised Contrastive Learning