paper-with-me

홈 › Papers

Long-tail learning via logit adjustment

2020-07-14 · ICLR 2021 1 · Aditya Krishna Menon, Sadeep Jayasumana, Ankit Singh Rawat, Himanshu Jain, Andreas Veit, Sanjiv Kumar

Real-world classification problems typically exhibit an imbalanced or long-tailed label distribution, wherein many labels are associated with only a few samples. This poses a challenge for generalisation on such labels, and also makes na\"ive learning biased towards dominant labels. In this paper, we present two simple modifications of standard softmax cross-entropy training to cope with these challenges. Our techniques revisit the classic idea of logit adjustment based on the label frequencies, either applied post-hoc to a trained model, or enforced in the loss during training. Such adjustment encourages a large relative margin between logits of rare versus dominant labels. These techniques unify and generalise several recent proposals in the literature, while possessing firmer statistical grounding and empirical performance.

📄 PDF Abstract BibTeX arXiv:2007.07314

Code (3)

google-research/google-research/tree/master/logit_adjustment 공식 구현 tf
Chumsy0725/logit-adj-pytorch pytorch
hbzju/solar pytorch

Tasks

Long-tail Learning

Methods 이 논문이 사용한 방법론

Softmax The Softmax output function transforms a previous layer's output into a vector of probabilities. It is commonly used for multiclass classification. Given an input vector $x$…

Similar Papers 제목 키워드 기반

Local and Global Logit Adjustments for Long-Tailed Learning

2023-01-01 · ICCV 2023 1 · Yingfan Tao, Jingna Sun, Hao Yang, Li Chen 외

Multi-expert ensemble models for long-tailed learning typically either learn diverse generalists from the whole dataset or aggregate specialists on different subsets. However, the former is insufficient for tail clas…

Long-Tail Temporal Action Segmentation with Group-wise Temporal Logit Adjustment

2024-08-19 · Zhanzhong Pang, Fadime Sener, Shrinivas Ramasubramanian, Angela Yao

Procedural activity videos often exhibit a long-tailed action distribution due to varying action frequencies and durations. However, state-of-the-art temporal action segmentation methods overlook the long tail and fail t…

Action SegmentationSegmentationTemporal Action Segmentation

Adaptive Logit Adjustment Loss for Long-Tailed Visual Recognition

2021-04-13 · Yan Zhao, Weicong Chen, Xu Tan, Kai Huang 외

Data in the real world tends to exhibit a long-tailed label distribution, which poses great challenges for the training of neural networks in visual recognition. Existing methods tackle this problem mainly from the persp…

General ClassificationSemantic SimilaritySemantic Textual Similarity

Long-tailed Visual Recognition via Gaussian Clouded Logit Adjustment

2023-05-19 · CVPR 2022 1 · Mengke Li, Yiu-ming Cheung, Yang Lu

Long-tailed data is still a big challenge for deep neural networks, even though they have achieved great success on balanced data. We observe that vanilla training on long-tailed data with cross-entropy loss makes the in…

Gradient-Aware Logit Adjustment Loss for Long-tailed Classifier

2024-03-14 · Fan Zhang, Wei Qin, Weijieying Ren, Lei Wang 외

In the real-world setting, data often follows a long-tailed distribution, where head classes contain significantly more training samples than tail classes. Consequently, models trained on such data tend to be biased towa…