Title:
MACHINE LEARNING WITH INSTANCE-DEPENDENT LABEL NOISE
Document Type and Number:
WIPO Patent Application WO/2023/153872
Kind Code:
A1
Abstract:
An artificial intelligence (AI) classifier is trained using supervised training and an effect of noise in the training data is reduced. The training data includes observed noisy labels. A posterior transition matrix (PTM) is used to minimize, in a statistical sense, a cross entropy between a noisy label and a function of the classifier output. A loss function using the PTM is provided to use in training the classifier. The classifier provides final output predictions with good performance even with the existence of noisy labels. Also, information fusion is included in the classifier training using the PTM and an estimated noise transition matrix (NTM) to reduce estimation error at the classifier output.
Inventors:
JIANG ZHIMENG (US)
ZHOU KAIXIONG (US)
LIU ZIRUI (US)
LI LI (US)
CHEN RUI (US)
CHOI SOO-HYUN (US)
HU XIA (US)
ZHOU KAIXIONG (US)
LIU ZIRUI (US)
LI LI (US)
CHEN RUI (US)
CHOI SOO-HYUN (US)
HU XIA (US)
Application Number:
PCT/KR2023/002015
Publication Date:
August 17, 2023
Filing Date:
February 10, 2023
Export Citation:
Assignee:
SAMSUNG ELECTRONICS CO LTD (KR)
International Classes:
G06N3/09; G06N3/045
Foreign References:
| CN111814962A | 2020-10-23 | |||
| US20200134454A1 | 2020-04-30 |
Other References:
YIVAN ZHANG; GANG NIU; MASASHI SUGIYAMA: "Learning Noise Transition Matrix from Only Noisy Labels via Total Variation Regularization", ARXIV.ORG, CORNELL UNIVERSITY LIBRARY, 201 OLIN LIBRARY CORNELL UNIVERSITY ITHACA, NY 14853, 14 June 2021 (2021-06-14), 201 Olin Library Cornell University Ithaca, NY 14853 , XP081978924
SEONG MIN KYE; KWANGHEE CHOI; JOONYOUNG YI; BURU CHANG: "Learning with Noisy Labels by Efficient Transition Matrix Estimation to Combat Label Miscorrection", ARXIV.ORG, CORNELL UNIVERSITY LIBRARY, 201 OLIN LIBRARY CORNELL UNIVERSITY ITHACA, NY 14853, 29 November 2021 (2021-11-29), 201 Olin Library Cornell University Ithaca, NY 14853, XP091105089
PATRINI GIORGIO, ROZZA ALESSANDRO, MENON ADITYA KRISHNA, NOCK RICHARD, QU LIZHEN: "Making Deep Neural Networks Robust to Label Noise: A Loss Correction Approach", ARXIV.ORG, 22 March 2017 (2017-03-22), XP093084799
ZHIMENG JIANG ET AL.: "AN INFORMATION FUSION APPROACH TO LEARNING WITH INSTANCE-DEPENDENT LABEL NOISE", IN: TENTH INTERNATIONAL CONFERENCE ON LEARNING REPRESENTATIONS (ICLR 2022), 25 April 2022, pages 1 - 20
SEONG MIN KYE; KWANGHEE CHOI; JOONYOUNG YI; BURU CHANG: "Learning with Noisy Labels by Efficient Transition Matrix Estimation to Combat Label Miscorrection", ARXIV.ORG, CORNELL UNIVERSITY LIBRARY, 201 OLIN LIBRARY CORNELL UNIVERSITY ITHACA, NY 14853, 29 November 2021 (2021-11-29), 201 Olin Library Cornell University Ithaca, NY 14853, XP091105089
PATRINI GIORGIO, ROZZA ALESSANDRO, MENON ADITYA KRISHNA, NOCK RICHARD, QU LIZHEN: "Making Deep Neural Networks Robust to Label Noise: A Loss Correction Approach", ARXIV.ORG, 22 March 2017 (2017-03-22), XP093084799
ZHIMENG JIANG ET AL.: "AN INFORMATION FUSION APPROACH TO LEARNING WITH INSTANCE-DEPENDENT LABEL NOISE", IN: TENTH INTERNATIONAL CONFERENCE ON LEARNING REPRESENTATIONS (ICLR 2022), 25 April 2022, pages 1 - 20
Attorney, Agent or Firm:
KIM, Tae-hun et al. (KR)
Download PDF: