본문 바로가기

쓰기

 
저자 김누리, 이지형 
학회명 IEEE/CVF Conference on Computer Vision and Pattern Recognition 2022 
학회명 (약자) CVPR 2022 
pp.  
학회시작일 2022-06-21 
학회종료일 2022-06-24 
비고  

Propagation Regularizer for Semi-supervised Learning with Extremely Scarce Labeled Samples


Abstract : Semi-supervised learning (SSL) is a method to make better models using a large number of easily accessible unlabeled data along with a small number of labeled data obtained at a high cost. Most of existing SSL studies focused on the cases where sufficient labeled samples were available, tens to hundreds labeled samples for class, which still requires a lot of labeling cost. In this paper, we focus on the SSL environment with extremely scarce labeled samples, 1 or 2 labeled samples per class, where most of existing methods failed to learn. We propose a propagation regularizer which can achieve efficient and effective learning with extremely scarce labeled samples by suppressing confirmation bias. In addition, for the realistic model selection in the absence of the validation dataset, we also propose a model selection method based on the propagation regularization in the SSL environment. The proposed methods show 70.9%, 30.3%, and 78.9% accuracy on CIFAR-10, CIFAR100, SVHN dataset with just one labeled sample per class, which are improved by 8.9% to 120.2% compared to the existing approaches.

    2024

      Learning with Structural Labels for Learning with Noisy Labels
      2024.02.27
      저자: 김누리*, 이진섭*, 이지형 (*: equal contribution)     학회명: IEEE/CVF Conference on Computer Vision and Pattern Recognition 2024     학회명 (약자): CVPR     학회시작일: 2024-06-17     학회종료일: 2024-06-19    

    2023

    2022

      Reducing computational cost in federated ensemble learning via rank-one matrix
      2022.12.05
      저자: YongHoon Kang, HoSeung Kim, Jee-Hyong Lee     학회명: Joint 12th International Conference on Soft Computing and Intelligent Systems and 23rd International Symposium on Advanced Intelligent Systems     학회명 (약자): SCIS-ISIS 2022     학회시작일: 2022-11-29     학회종료일: 2022-12-02    

    2021

      Class Activation Map Pairing for Adversarial Robustness
      2021.01.24
      저자: DongEon Jeong, MinGyu Lee, JinSeob Lee, Jee-Hyong Lee     학회명: International Conference on Convergence Content 2020     학회명 (약자): ICCC     pp.: 419-420     학회시작일: 2020-11-16     학회종료일: 2020-11-18