Timezone: »

Large-Margin Contrastive Learning with Distance Polarization Regularizer
Shuo Chen · Gang Niu · Chen Gong · Jun Li · Jian Yang · Masashi Sugiyama

Tue Jul 20 09:00 AM -- 11:00 AM (PDT) @

\emph{Contrastive learning}~(CL) pretrains models in a pairwise manner, where given a data point, other data points are all regarded as dissimilar, including some that are \emph{semantically} similar. The issue has been addressed by properly weighting similar and dissimilar pairs as in \emph{positive-unlabeled learning}, so that the objective of CL is \emph{unbiased} and CL is \emph{consistent}. However, in this paper, we argue that this great solution is still not enough: its weighted objective \emph{hides} the issue where the semantically similar pairs are still pushed away; as CL is pretraining, this phenomenon is not our desideratum and might affect downstream tasks. To this end, we propose \emph{large-margin contrastive learning}~(LMCL) with \emph{distance polarization regularizer}, motivated by the distribution characteristic of pairwise distances in \emph{metric learning}. In LMCL, we can distinguish between \emph{intra-cluster} and \emph{inter-cluster} pairs, and then only push away inter-cluster pairs, which \emph{solves} the above issue explicitly. Theoretically, we prove a tighter error bound for LMCL; empirically, the superiority of LMCL is demonstrated across multiple domains, \emph{i.e.}, image classification, sentence representation, and reinforcement learning.

Author Information

Shuo Chen (RIKEN)
Gang Niu (RIKEN)
Gang Niu

Gang Niu is currently an indefinite-term senior research scientist at RIKEN Center for Advanced Intelligence Project.

Chen Gong (Nanjing University of Science and Technology)
Jun Li (Nanjing University of Science and Technology)
Jian Yang (Nanjing University of Science and Technology)
Masashi Sugiyama (RIKEN / The University of Tokyo)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors