paper-with-me

홈 › Papers

Dual Temperature Helps Contrastive Learning Without Many Negative Samples: Towards Understanding and Simplifying MoCo

2022-03-30 · CVPR 2022 1 · Chaoning Zhang, Kang Zhang, Trung X. Pham, Axi Niu, Zhinan Qiao, Chang D. Yoo, In So Kweon

Contrastive learning (CL) is widely known to require many negative samples, 65536 in MoCo for instance, for which the performance of a dictionary-free framework is often inferior because the negative sample size (NSS) is limited by its mini-batch size (MBS). To decouple the NSS from the MBS, a dynamic dictionary has been adopted in a large volume of CL frameworks, among which arguably the most popular one is MoCo family. In essence, MoCo adopts a momentum-based queue dictionary, for which we perform a fine-grained analysis of its size and consistency. We point out that InfoNCE loss used in MoCo implicitly attract anchors to their corresponding positive sample with various strength of penalties and identify such inter-anchor hardness-awareness property as a major reason for the necessity of a large dictionary. Our findings motivate us to simplify MoCo v2 via the removal of its dictionary as well as momentum. Based on an InfoNCE with the proposed dual temperature, our simplified frameworks, SimMoCo and SimCo, outperform MoCo v2 by a visible margin. Moreover, our work bridges the gap between CL and non-CL frameworks, contributing to a more unified understanding of these two mainstream frameworks in SSL. Code is available at: https://bit.ly/3LkQbaT.

📄 PDF Abstract BibTeX arXiv:2203.17248

Code (2)

ChaoningZhang/Dual-temperature 공식 구현 pytorch
pantheon5100/deacl pytorch

Tasks

Contrastive Learning

Methods 이 논문이 사용한 방법론

Random Gaussian Blur Random Gaussian Blur is an image data augmentation technique where we randomly blur the image using a Gaussian distribution. Image Source:…
Dense Connections Dense Connections, or Fully Connected Connections, are a type of layer in a deep neural network that use a linear operation where every input is connected to every output…
InfoNCE 설명 없음
Feedforward Network A Feedforward Network, or a Multilayer Perceptron (MLP), is a neural network with solely densely connected layers. This is the classic neural network architecture of the…
Batch Normalization 설명 없음
MoCo v2 MoCo v2 is an improved version of the Momentum Contrast self-supervised learning algorithm. Motivated by the findings presented in…
MoCo 설명 없음

Similar Papers 제목 키워드 기반

Not All Semantics are Created Equal: Contrastive Self-supervised Learning with Automatic Temperature Individualization

2023-05-19 · Zi-Hao Qiu, Quanqi Hu, Zhuoning Yuan, Denny Zhou 외

In this paper, we aim to optimize a contrastive loss with individualized temperatures in a principled and systematic manner for self-supervised learning. The common practice of using a global temperature parameter $\tau$…

AllSelf-Supervised Learning

Temperature as Uncertainty in Contrastive Learning

2021-10-08 · Oliver Zhang, Mike Wu, Jasmine Bayrooti, Noah Goodman

Contrastive learning has demonstrated great capability to learn representations without annotations, even outperforming supervised baselines. However, it still lacks important properties useful for real-world application…

Contrastive LearningLinear evaluationOut-of-Distribution Detection

On the modality gap and the contrastive loss in multi-modal representation learning

2026-07-12 · Fabian Mager, Hiba Nassar, Lars Kai Hansen arxiv

We study the modality gap in CLIP-style dual-encoder contrastive learning, where image and text embeddings remain misaligned despite being trained in a shared space. We argue that the gap is induced by a failure of the I…

Representation LearningContrastive Learning

Understanding the Behaviour of Contrastive Loss

2020-12-15 · CVPR 2021 1 · Feng Wang, Huaping Liu

Unsupervised contrastive learning has achieved outstanding success, while the mechanism of contrastive loss has been less studied. In this paper, we concentrate on the understanding of the behaviours of unsupervised cont…

Contrastive Learning

Dual-level Hypergraph Contrastive Learning with Adaptive Temperature Enhancement

2024-05-14 · International World Wide Web Conference 2024 5 · Yiyue Qian, Tianyi Ma, Chuxu Zhang, Yanfang Ye

Inspired by the success of graph contrastive learning, researchers have begun exploring the benefits of contrastive learning over hypergraphs. However, these works have the following limitations in modeling the high-orde…

Contrastive LearningHypergraph Contrastive LearningHypergraph representationsRepresentation Learning