Temperature-Free Loss Function for Contrastive Learning
Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 reserved pith:SHD45MHIrecord.jsonopen to challenge →
read the original abstract
As one of the most promising methods in self-supervised learning, contrastive learning has achieved a series of breakthroughs across numerous fields. A predominant approach to implementing contrastive learning is applying InfoNCE loss: By capturing the similarities between pairs, InfoNCE loss enables learning the representation of data. Albeit its success, adopting InfoNCE loss requires tuning a temperature, which is a core hyperparameter for calibrating similarity scores. Despite its significance and sensitivity to performance being emphasized by several studies, searching for a valid temperature requires extensive trial-and-error-based experiments, which increases the difficulty of adopting InfoNCE loss. To address this difficulty, we propose a novel method to deploy InfoNCE loss without temperature. Specifically, we replace temperature scaling with the inverse hyperbolic tangent function, resulting in a modified InfoNCE loss. In addition to hyperparameter-free deployment, we observed that the proposed method even yielded a performance gain in contrastive learning. Our detailed theoretical analysis discovers that the current practice of temperature scaling in InfoNCE loss causes serious problems in gradient descent, whereas our method provides desirable gradient properties. The proposed method was validated on five benchmarks on contrastive learning, yielding satisfactory results without temperature tuning.
This paper has not been read by Pith yet.
Forward citations
Cited by 2 Pith papers
-
When Softmax Fails at the Top: Extreme Value Corrections for InfoNCE
InfoNCE softmax misaligned with normalized embeddings per extreme value theory; WEINCE adds batch-based endpoint shortfall correction for consistent gains on five vision benchmarks.
-
MaCo-GAN: Manifold-Contrastive Adversarial Learning for Single Image Super-Resolution
MaCo-GAN introduces a manifold-contrastive GAN that replaces adversarial loss with a contrastive minimax game over synthesized fake samples to improve the perception-distortion trade-off in SISR.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.