| 발표자 | 김누리 |
|---|---|
| 발표일자 | 2022-10-05 |
| 저자 | Wang, Xudong, et al. |
| 학회명 | Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2022. |
| 논문지 |
This work studies the bias issue of pseudo-labeling, a natural phenomenon that widely occurs but often overlooked by prior research. Pseudo-labels are generated when a classifier trained on source data is transferred to unlabeled target data. We observe heavy long-tailed pseudo-labels when a semi-supervised learning model FixMatch predicts labels on the unlabeled set even though the unlabeled data is curated to be balanced. Without intervention, the training model inherits the bias from the pseudo-labels and end up being sub-optimal. To eliminate the model bias, we propose a simple yet effective method DebiasMatch, comprising of an adaptive debiasing module and an adaptive marginal loss. The strength of debiasing and the size of margins can be automatically adjusted by making use of an online updated queue. Benchmarked on ImageNet-1K, DebiasMatch significantly outperforms previous state-of-the-arts by more than 26% and 10.5% on semi-supervised learning (0.2% annotated data) and zero-shot learning tasks respectively.
댓글 0
2023
Querying Labeled for Unlabeled: Cross-Image Semantic Consistency Guided Semi-Supervised Semantic Segmentation
2023.02.22
발표자: 임지영
발표일자: 2023-02-22
학회명: IEEE Transactions on Pattern Analysis and Machine Intelligence
Neural Tangent Kernel:Convergence and Generalization in Neural Networks
2023.02.22
발표자: 안재한
발표일자: 2023-02-22
저자: Arthur Jacot, Franck Gabriel, Clement Hongler
학회명: NeurIPS 2018
Toward Deep Supervised Anomaly Detection Reinforcement Learning from Partially Labeled Anomaly Data
2023.02.15
발표자: 장성인
발표일자: 2023-02-15
저자: Guansong Pang et al
학회명: Knowledge Discovery and Data Mining (KDD) 2021
NUWA : Visual Synthesis Pre-training for Neural visUal World creAtion
2023.02.15
발표자: 송경렬
발표일자: 2023-02-15
저자: Chenfei Wu1* Jian Liang2* Lei Ji1 Fan Yang1 Yuejian Fang2 Daxin Jiang1 Nan Duan1†
Improving Contrastive Learning by Visualizing Feature Transformation
2023.02.15
발표자: 이진섭
발표일자: 2023-02-15
저자: Rui Zhu, Bingchen Zhao, Jingen Liu , Zhenglong Sun, Chang Wen Chen
학회명: ICCV 2021 Oral
2022
Why Should Adversarial Perturbations be Imperceptible? Rethink the Research Paradigm in Adversarial NLP
2022.12.29
발표자: 김효준
발표일자: 2023-01-11
저자: Yangyi Chen, Hongcheng Gao, Ganqu Cui, Fanchao Qi, Longtao Huang, Zhiyuan Liu, Maosong Sun
학회명: EMNLP 2022
TAGE: Task-Agnostic Graph Explainers
2023.01.11
발표자: 지혜림
발표일자: 2023-01-11
저자: Yaochen Xie, Sumeet Katariya, Xianfeng Tang, Edward W Huang, Nikhil Rao, Karthik Subbian, Shuiwang Ji
학회명: NeurIPS 2022
Threshold Matters in WSSS: Manipulating the Activation for the Robust and Accurate Segmentation Model Against Thresholds
2023.01.11
발표자: 최준수
발표일자: 2023-01-11
저자: Minhyun Lee, Dongseob Kim, Hyunjung Shim
학회명: CVPR 2022
Promoting Semantic Connectivity : Dual Nearest Neighbors Contrastive Learning for Unsupervised Domain Generalization
2022.12.21
발표자: 이진섭
발표일자: 2022-12-22
저자: Yuchen Liu, Yaoming Wang, Yabo Chen, Wenrui Dai, Chenglin Li, Junni Zou, Hongkai Xiong
Self-distilled StyleGAN: Towards Generation from Internet Photos
2022.12.22
발표자: 홍만수
발표일자: 2022-12-22
저자: Ron Mokady, Michal Yarom, Omer Tov, Oran Lang, Daniel Cohen-Or, Tali Dekel, Michal Irani, Inbar Mosseri
PGExplainer: Parameterized Explainer for Graph Neural Network
2022.12.14
발표자: 지혜림
발표일자: 2022-12-15
저자: Dongsheng Luo, Wei Cheng, Dongkuan Xu, Wenchao Yu, Bo Zong, Haifeng Chen, Xiang Zhang
학회명: NeurIPS 2020
Template-free Prompt Tuning for Few-shot NER
2022.12.15
발표자: 안지민
발표일자: 2022-12-15
저자: Ruotian Ma1, Xin Zhou1, Tao Gui2, Yiding Tan1, Linyang Li1, Qi Zhang1, Xuanjing Huang1
학회명: NAACL 2021
Unsupervised Domain Generalization by Learning a Bridge Across Domains
2022.12.15
발표자: 김누리
발표일자: 2022-12-15
저자: Harary, Sivan, et al.
학회명: CVPR 2022
Domain-Oriented Prefix-Tuning: Towards Efficient and Generalizable Fine-tuning for Zero-Shot Dialogue Summarization
2022.11.23
발표자: 최윤석
발표일자: 2022-11-23
저자: Lulu Zhao, Fujia Zheng, Weihao Zeng, Keqing He, Weiran Xu, Huixing Jiang, Wei Wu, Yanan Wu
학회명: NAACL 2022
Selective-Supervised Contrastive Learning with Noisy Labels
2022.11.23
발표자: 송경렬
발표일자: 2022-11-23
저자: Shikun Li1,2, Xiaobo Xia3, Shiming Ge1,2*, Tongliang Liu3
학회명: CVPR 2022
Class Re-Activation Maps for Weakly-Supervised Semantic Segmentation
2022.11.09
발표자: 최준수
발표일자: 2022-11-09
저자: Zhaozheng Chen, Tan Wang, Xiongwei Wu et al.
학회명: CVPR 2022
Acknowledging the Unknown for Multi-label Learning with Single Positive Labels
2022.11.09
발표자: 송경렬
발표일자: 2022-11-09
저자: Donghao Zhou, Pengfei Chen, Qiong Wang, Guangyong Chen, and Pheng-Ann Heng
학회명: ECCV 2022
Sharpness-Aware Minimization for Efficiently Improving Generalization
2022.11.09
발표자: 안재한
발표일자: 2022-11-09
저자: Pierre Foret, Ariel Kleiner, Hossein Mobahi, and Behnam Neyshabur
학회명: ICLR 2021