Classification with Noisy Labels by Importance Reweighting
In this paper, we study a classification problem in which sample labels are randomly corrupted. In this scenario, there is an unobservable sample with noise-free labels. However, before being observed, the true labels are independently flipped with a probability <inline-formula><tex-math>...
Uloženo v:
| Vydáno v: | IEEE transactions on pattern analysis and machine intelligence Ročník 38; číslo 3; s. 447 - 461 |
|---|---|
| Hlavní autoři: | , |
| Médium: | Journal Article |
| Jazyk: | angličtina |
| Vydáno: |
United States
IEEE
01.03.2016
The Institute of Electrical and Electronics Engineers, Inc. (IEEE) |
| Témata: | |
| ISSN: | 0162-8828, 1939-3539, 2160-9292, 1939-3539 |
| On-line přístup: | Získat plný text |
| Tagy: |
Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
|
| Abstract | In this paper, we study a classification problem in which sample labels are randomly corrupted. In this scenario, there is an unobservable sample with noise-free labels. However, before being observed, the true labels are independently flipped with a probability <inline-formula><tex-math>\rho \in [0,0.5)</tex-math> <inline-graphic xlink:type="simple" xlink:href="tao-ieq1-2456899.gif"/> </inline-formula>, and the random label noise can be class-conditional. Here, we address two fundamental problems raised by this scenario. The first is how to best use the abundant surrogate loss functions designed for the traditional classification problem when there is label noise. We prove that any surrogate loss function can be used for classification with noisy labels by using importance reweighting, with consistency assurance that the label noise does not ultimately hinder the search for the optimal classifier of the noise-free sample. The other is the open problem of how to obtain the noise rate <inline-formula> <tex-math>\rho</tex-math> <inline-graphic xlink:type="simple" xlink:href="tao-ieq2-2456899.gif"/> </inline-formula>. We show that the rate is upper bounded by the conditional probability <inline-formula><tex-math> P(\hat{Y}|X)</tex-math> <inline-graphic xlink:type="simple" xlink:href="tao-ieq3-2456899.gif"/> </inline-formula> of the noisy sample. Consequently, the rate can be estimated, because the upper bound can be easily reached in classification problems. Experimental results on synthetic and real datasets confirm the efficiency of our methods. |
|---|---|
| AbstractList | In this paper, we study a classification problem in which sample labels are randomly corrupted. In this scenario, there is an unobservable sample with noise-free labels. However, before being observed, the true labels are independently flipped with a probability ρ ∈ [0,0.5), and the random label noise can be class-conditional. Here, we address two fundamental problems raised by this scenario. The first is how to best use the abundant surrogate loss functions designed for the traditional classification problem when there is label noise. We prove that any surrogate loss function can be used for classification with noisy labels by using importance reweighting, with consistency assurance that the label noise does not ultimately hinder the search for the optimal classifier of the noise-free sample. The other is the open problem of how to obtain the noise rate ρ. We show that the rate is upper bounded by the conditional probability P(∧Y|X) of the noisy sample. Consequently, the rate can be estimated, because the upper bound can be easily reached in classification problems. Experimental results on synthetic and real datasets confirm the efficiency of our methods.In this paper, we study a classification problem in which sample labels are randomly corrupted. In this scenario, there is an unobservable sample with noise-free labels. However, before being observed, the true labels are independently flipped with a probability ρ ∈ [0,0.5), and the random label noise can be class-conditional. Here, we address two fundamental problems raised by this scenario. The first is how to best use the abundant surrogate loss functions designed for the traditional classification problem when there is label noise. We prove that any surrogate loss function can be used for classification with noisy labels by using importance reweighting, with consistency assurance that the label noise does not ultimately hinder the search for the optimal classifier of the noise-free sample. The other is the open problem of how to obtain the noise rate ρ. We show that the rate is upper bounded by the conditional probability P(∧Y|X) of the noisy sample. Consequently, the rate can be estimated, because the upper bound can be easily reached in classification problems. Experimental results on synthetic and real datasets confirm the efficiency of our methods. In this paper, we study a classification problem in which sample labels are randomly corrupted. In this scenario, there is an unobservable sample with noise-free labels. However, before being observed, the true labels are independently flipped with a probability <inline-formula><tex-math>\rho \in [0,0.5)</tex-math> <inline-graphic xlink:type="simple" xlink:href="tao-ieq1-2456899.gif"/> </inline-formula>, and the random label noise can be class-conditional. Here, we address two fundamental problems raised by this scenario. The first is how to best use the abundant surrogate loss functions designed for the traditional classification problem when there is label noise. We prove that any surrogate loss function can be used for classification with noisy labels by using importance reweighting, with consistency assurance that the label noise does not ultimately hinder the search for the optimal classifier of the noise-free sample. The other is the open problem of how to obtain the noise rate <inline-formula> <tex-math>\rho</tex-math> <inline-graphic xlink:type="simple" xlink:href="tao-ieq2-2456899.gif"/> </inline-formula>. We show that the rate is upper bounded by the conditional probability <inline-formula><tex-math> P(\hat{Y}|X)</tex-math> <inline-graphic xlink:type="simple" xlink:href="tao-ieq3-2456899.gif"/> </inline-formula> of the noisy sample. Consequently, the rate can be estimated, because the upper bound can be easily reached in classification problems. Experimental results on synthetic and real datasets confirm the efficiency of our methods. In this paper, we study a classification problem in which sample labels are randomly corrupted. In this scenario, there is an unobservable sample with noise-free labels. However, before being observed, the true labels are independently flipped with a probability ρ ∈ [0,0.5), and the random label noise can be class-conditional. Here, we address two fundamental problems raised by this scenario. The first is how to best use the abundant surrogate loss functions designed for the traditional classification problem when there is label noise. We prove that any surrogate loss function can be used for classification with noisy labels by using importance reweighting, with consistency assurance that the label noise does not ultimately hinder the search for the optimal classifier of the noise-free sample. The other is the open problem of how to obtain the noise rate ρ. We show that the rate is upper bounded by the conditional probability P(∧Y|X) of the noisy sample. Consequently, the rate can be estimated, because the upper bound can be easily reached in classification problems. Experimental results on synthetic and real datasets confirm the efficiency of our methods. In this paper, we study a classification problem in which sample labels are randomly corrupted. In this scenario, there is an unobservable sample with noise-free labels. However, before being observed, the true labels are independently flipped with a probability [Formula Omitted], and the random label noise can be class-conditional. Here, we address two fundamental problems raised by this scenario. The first is how to best use the abundant surrogate loss functions designed for the traditional classification problem when there is label noise. We prove that any surrogate loss function can be used for classification with noisy labels by using importance reweighting, with consistency assurance that the label noise does not ultimately hinder the search for the optimal classifier of the noise-free sample. The other is the open problem of how to obtain the noise rate [Formula Omitted]. We show that the rate is upper bounded by the conditional probability [Formula Omitted] of the noisy sample. Consequently, the rate can be estimated, because the upper bound can be easily reached in classification problems. Experimental results on synthetic and real datasets confirm the efficiency of our methods. |
| Author | Liu, Tongliang Tao, Dacheng |
| Author_xml | – sequence: 1 givenname: Tongliang surname: Liu fullname: Liu, Tongliang email: tliang.liu@gmail.com organization: Centre for Quantum Computation & Intelligent Systems and the Faculty of Engineering and Information Technology, University of Technology, Sydney, 81 Broadway Street, Ultimo, NSW 2007, Australia – sequence: 2 givenname: Dacheng surname: Tao fullname: Tao, Dacheng email: dacheng.tao@uts.edu.au organization: Centre for Quantum Computation & Intelligent Systems and the Faculty of Engineering and Information Technology, University of Technology, Sydney, 81 Broadway Street, Ultimo, NSW 2007, Australia |
| BackLink | https://www.ncbi.nlm.nih.gov/pubmed/27046490$$D View this record in MEDLINE/PubMed |
| BookMark | eNp9kMlOwzAQQC1URMvyAyChSFy4pHiLY3OrKpZKZRGCc2Q7Y-oqS4kTof49KS0cOHDy5b2Z8TtEg6quAKFTgseEYHX1-jx5mI0pJsmY8kRIpfbQiBKBY0UVHaARJoLGUlI5RIchLDEmPMHsAA1pirngCo_Q9bTQIXjnrW59XUWfvl1Ej7UP62iuDRQhMutoVq7qptWVhegFPsG_L1pfvR-jfaeLACe79wi93d68Tu_j-dPdbDqZx5YlpI0tVc5yYE4rKRQz3CXUyFwzh1Ojmci1A6WtTAFyK4RlxnJDnHGA85xix47Q5Xbuqqk_OghtVvpgoSh0BXUXMpKmivc_E6xHL_6gy7prqv66DSU5TURCeup8R3WmhDxbNb7UzTr7qdIDdAvYpg6hAfeLEJxt0mff6bNN-myXvpfkH8n69jtq22hf_K-ebVUPAL-7UpIogjH7AmBjkTc |
| CODEN | ITPIDJ |
| CitedBy_id | crossref_primary_10_1016_j_neucom_2018_11_009 crossref_primary_10_1088_1361_6560_ad4083 crossref_primary_10_1109_TNNLS_2016_2598744 crossref_primary_10_1016_j_neucom_2019_05_019 crossref_primary_10_1109_ACCESS_2019_2918794 crossref_primary_10_1007_s11042_018_6277_x crossref_primary_10_1109_ACCESS_2017_2767907 crossref_primary_10_1007_s11081_025_10014_z crossref_primary_10_1016_j_ins_2023_119644 crossref_primary_10_1007_s11263_024_02205_5 crossref_primary_10_1109_TPAMI_2024_3462996 crossref_primary_10_1007_s11063_017_9721_4 crossref_primary_10_1109_LGRS_2017_2764042 crossref_primary_10_1155_2018_5381962 crossref_primary_10_1109_TIP_2024_3374221 crossref_primary_10_1016_j_knosys_2023_110819 crossref_primary_10_1016_j_epsr_2019_03_009 crossref_primary_10_1109_TCYB_2016_2533430 crossref_primary_10_1016_j_neunet_2025_107464 crossref_primary_10_1109_TCYB_2018_2816984 crossref_primary_10_1109_TKDE_2021_3049540 crossref_primary_10_1016_j_fmre_2022_09_019 crossref_primary_10_1109_TIP_2015_2495116 crossref_primary_10_1088_1742_6596_2433_1_012039 crossref_primary_10_1002_mp_70007 crossref_primary_10_1109_TIP_2016_2598654 crossref_primary_10_1109_TNNLS_2022_3190528 crossref_primary_10_1109_TMM_2016_2608780 crossref_primary_10_1109_TPAMI_2024_3360623 crossref_primary_10_1109_TETCI_2019_2917704 crossref_primary_10_3390_rs12182957 crossref_primary_10_1109_TNNLS_2023_3271140 crossref_primary_10_1007_s11263_018_1075_5 crossref_primary_10_1109_TIP_2016_2549360 crossref_primary_10_1109_TNNLS_2020_3009417 crossref_primary_10_1109_TPAMI_2017_2772235 crossref_primary_10_1016_j_neunet_2024_107041 crossref_primary_10_1109_JSTARS_2020_2994162 crossref_primary_10_1109_TIP_2016_2570553 crossref_primary_10_1002_sim_10315 crossref_primary_10_1145_2943785 crossref_primary_10_1016_j_patcog_2021_108198 crossref_primary_10_1109_TIP_2016_2613686 crossref_primary_10_1016_j_swevo_2017_11_003 crossref_primary_10_1109_TPAMI_2022_3180545 crossref_primary_10_1016_j_neunet_2020_02_010 crossref_primary_10_1109_TNNLS_2020_3047046 crossref_primary_10_1007_s11227_020_03480_y crossref_primary_10_1016_j_neucom_2022_01_031 crossref_primary_10_1109_TCYB_2019_2950560 crossref_primary_10_1109_TIE_2016_2552147 crossref_primary_10_1016_j_image_2017_05_001 crossref_primary_10_1016_j_eswa_2017_09_057 crossref_primary_10_1109_TNNLS_2017_2682179 crossref_primary_10_1007_s10994_020_05877_5 crossref_primary_10_1016_j_neunet_2024_107044 crossref_primary_10_1109_TMI_2022_3219126 crossref_primary_10_1109_TNNLS_2021_3111871 crossref_primary_10_1007_s00138_018_0941_z crossref_primary_10_1109_TGRS_2018_2861992 crossref_primary_10_1016_j_ins_2016_02_030 crossref_primary_10_1007_s10994_021_06081_9 crossref_primary_10_1038_s41598_021_84499_w crossref_primary_10_1109_TPAMI_2022_3178690 crossref_primary_10_1109_LSP_2024_3496584 crossref_primary_10_1109_TIP_2018_2877939 crossref_primary_10_1109_TMM_2024_3390113 crossref_primary_10_1109_TPAMI_2025_3536845 crossref_primary_10_3390_rs13214194 crossref_primary_10_1109_TCYB_2016_2626295 crossref_primary_10_1145_3749177 crossref_primary_10_1016_j_patcog_2019_01_033 crossref_primary_10_1016_j_artint_2025_104363 crossref_primary_10_1016_j_patcog_2022_109210 crossref_primary_10_3390_rs15102543 crossref_primary_10_1109_TNNLS_2021_3105984 crossref_primary_10_1016_j_knosys_2025_113783 crossref_primary_10_1007_s10115_020_01466_z crossref_primary_10_1109_TIFS_2021_3058771 crossref_primary_10_1109_TPAMI_2022_3141240 crossref_primary_10_1109_ACCESS_2025_3526332 crossref_primary_10_1007_s11042_016_3370_x crossref_primary_10_1007_s00357_023_09430_6 crossref_primary_10_1109_JSTARS_2022_3230625 crossref_primary_10_1145_3365672 crossref_primary_10_1016_j_foodres_2025_116741 crossref_primary_10_3233_JIFS_200432 crossref_primary_10_1007_s11704_021_1208_9 crossref_primary_10_1631_FITEE_2200514 crossref_primary_10_1080_01621459_2021_2016423 crossref_primary_10_3389_fnbot_2022_1083835 crossref_primary_10_1109_TNNLS_2017_2699783 crossref_primary_10_1007_s11263_024_02012_y crossref_primary_10_1109_TIP_2016_2556940 crossref_primary_10_1186_s12859_015_0844_1 crossref_primary_10_1109_TCSVT_2016_2539778 crossref_primary_10_1109_TVCG_2023_3345340 crossref_primary_10_1007_s11063_020_10346_0 crossref_primary_10_1109_TCYB_2015_2496974 crossref_primary_10_1007_s11042_018_5703_4 crossref_primary_10_1109_TNNLS_2024_3417433 crossref_primary_10_1016_j_optlaseng_2023_107798 crossref_primary_10_1007_s10489_025_06594_z crossref_primary_10_1016_j_ins_2017_01_011 crossref_primary_10_1109_TPAMI_2017_2751607 crossref_primary_10_1016_j_neunet_2021_07_024 crossref_primary_10_1109_TNNLS_2019_2935384 crossref_primary_10_1016_j_inffus_2024_102702 crossref_primary_10_1109_TKDE_2023_3328645 crossref_primary_10_1109_TGRS_2015_2506612 crossref_primary_10_1109_TIP_2019_2933724 crossref_primary_10_1109_TPAMI_2021_3094662 crossref_primary_10_1145_2956556 crossref_primary_10_1109_TNNLS_2016_2643286 crossref_primary_10_3390_rs16193568 crossref_primary_10_1214_16_EJS1193 crossref_primary_10_3390_e21100924 crossref_primary_10_1109_TCSVT_2022_3220669 crossref_primary_10_1109_TIP_2015_2473662 crossref_primary_10_1109_TPAMI_2021_3114089 crossref_primary_10_1109_TVT_2021_3066210 crossref_primary_10_1109_JAS_2023_124083 crossref_primary_10_1016_j_patrec_2018_09_008 crossref_primary_10_1016_j_media_2020_101759 crossref_primary_10_1007_s11760_024_03591_6 crossref_primary_10_1109_ACCESS_2020_3022672 crossref_primary_10_1109_ACCESS_2021_3109989 crossref_primary_10_1007_s12559_016_9446_0 crossref_primary_10_1007_s40745_024_00552_1 crossref_primary_10_1007_s00521_019_04627_6 crossref_primary_10_1016_j_jvcir_2016_03_007 crossref_primary_10_1109_TCYB_2016_2616119 crossref_primary_10_1016_j_knosys_2021_106771 crossref_primary_10_1109_TIP_2016_2573581 crossref_primary_10_1109_TMM_2021_3137752 crossref_primary_10_1109_TPAMI_2023_3301876 crossref_primary_10_1109_TIP_2019_2902115 crossref_primary_10_1111_coin_12638 crossref_primary_10_1109_ACCESS_2020_2968934 crossref_primary_10_1007_s10115_025_02595_z crossref_primary_10_1016_j_artint_2021_103451 crossref_primary_10_1109_TPAMI_2024_3489217 crossref_primary_10_1007_s10845_020_01608_8 crossref_primary_10_1016_j_ins_2021_08_008 crossref_primary_10_1007_s11222_023_10283_7 crossref_primary_10_1109_TNNLS_2017_2734812 crossref_primary_10_1109_TPAMI_2023_3247629 crossref_primary_10_1109_TII_2022_3229130 crossref_primary_10_1109_JIOT_2016_2561962 crossref_primary_10_1007_s10707_020_00399_7 crossref_primary_10_1007_s11704_023_2709_5 crossref_primary_10_1007_s00530_025_01701_z crossref_primary_10_1109_TCYB_2023_3339629 crossref_primary_10_1016_j_patcog_2025_111681 crossref_primary_10_1155_2018_2179049 crossref_primary_10_1007_s42967_020_00084_4 crossref_primary_10_1109_ACCESS_2021_3085114 crossref_primary_10_1109_TIP_2017_2651372 crossref_primary_10_1016_j_patcog_2024_111132 crossref_primary_10_1109_TSMC_2020_3018102 crossref_primary_10_1016_j_neucom_2022_02_030 crossref_primary_10_1109_TCYB_2019_2953337 crossref_primary_10_1109_ACCESS_2020_3041873 crossref_primary_10_1109_TIP_2016_2564638 crossref_primary_10_1007_s10994_024_06678_w crossref_primary_10_1109_TCBB_2022_3224934 crossref_primary_10_1109_COMST_2022_3200740 crossref_primary_10_1080_00401706_2022_2078413 crossref_primary_10_1109_TCSVT_2022_3231887 crossref_primary_10_1145_3644391 crossref_primary_10_1016_j_future_2023_03_007 crossref_primary_10_1109_LGRS_2017_2755541 crossref_primary_10_1145_3627157 crossref_primary_10_1109_TPAMI_2023_3347850 crossref_primary_10_1109_TCAD_2022_3218239 crossref_primary_10_1109_TIP_2018_2881928 crossref_primary_10_1109_TMI_2021_3091178 crossref_primary_10_1109_TMM_2022_3181439 crossref_primary_10_1016_j_neunet_2021_06_012 crossref_primary_10_1109_JPROC_2020_2989782 crossref_primary_10_1109_TCSVT_2024_3371513 crossref_primary_10_1016_j_displa_2023_102370 crossref_primary_10_3233_IDA_230818 crossref_primary_10_1016_j_asoc_2024_111400 crossref_primary_10_1016_j_knosys_2025_113499 crossref_primary_10_1016_j_ins_2016_08_040 crossref_primary_10_3233_IDA_184024 crossref_primary_10_1007_s11063_018_9963_9 crossref_primary_10_1109_TPAMI_2023_3275249 crossref_primary_10_1109_TPAMI_2025_3584051 crossref_primary_10_1007_s11063_016_9542_x crossref_primary_10_1109_TNNLS_2018_2792062 crossref_primary_10_1016_j_sigpro_2017_09_002 crossref_primary_10_1109_TKDE_2017_2654250 crossref_primary_10_1109_TNNLS_2019_2899045 crossref_primary_10_1145_3342241 crossref_primary_10_1109_TIP_2021_3089942 crossref_primary_10_1109_TMM_2021_3115635 crossref_primary_10_1145_3687468 crossref_primary_10_1109_TIP_2016_2612883 crossref_primary_10_1109_TCYB_2016_2608906 crossref_primary_10_1109_TPAMI_2024_3355402 crossref_primary_10_1016_j_isprsjprs_2022_03_005 crossref_primary_10_1109_TNNLS_2022_3152527 crossref_primary_10_1109_TIP_2024_3430539 crossref_primary_10_1109_TKDE_2023_3313604 crossref_primary_10_1016_j_neunet_2025_107977 crossref_primary_10_1109_TMC_2025_3558874 crossref_primary_10_1109_ACCESS_2020_3019465 crossref_primary_10_1145_3429447 crossref_primary_10_1145_3665895 crossref_primary_10_1016_j_arcontrol_2018_09_003 crossref_primary_10_1016_j_jvcir_2016_10_001 crossref_primary_10_1007_s11263_024_02324_z crossref_primary_10_1016_j_neucom_2020_03_002 crossref_primary_10_3390_app122211406 crossref_primary_10_1109_TIP_2016_2614132 crossref_primary_10_1109_TIP_2017_2722224 crossref_primary_10_1109_TPAMI_2024_3357814 crossref_primary_10_1109_JSTARS_2022_3168799 crossref_primary_10_1109_TCYB_2017_2739423 crossref_primary_10_1109_TPAMI_2020_3044997 crossref_primary_10_1109_TPAMI_2023_3271451 crossref_primary_10_3390_math10111892 crossref_primary_10_1109_JBHI_2020_3008601 crossref_primary_10_1007_s11263_025_02494_4 crossref_primary_10_1016_j_engappai_2023_107714 crossref_primary_10_1109_TCSVT_2025_3526449 crossref_primary_10_1109_TIP_2016_2563981 crossref_primary_10_1007_s10115_021_01643_8 crossref_primary_10_1007_s10618_018_0592_8 crossref_primary_10_1016_j_eswa_2023_122506 crossref_primary_10_1016_j_neucom_2016_09_016 crossref_primary_10_1117_1_JEI_33_1_013055 crossref_primary_10_1016_j_imavis_2020_104000 crossref_primary_10_1109_TCSVT_2023_3321733 crossref_primary_10_1109_TMM_2022_3152567 crossref_primary_10_1016_j_patcog_2023_110072 crossref_primary_10_1145_3551636 crossref_primary_10_1109_TIP_2017_2771408 crossref_primary_10_1093_jamia_ocae108 crossref_primary_10_1016_j_neunet_2022_08_014 crossref_primary_10_1016_j_neucom_2016_10_073 crossref_primary_10_1016_j_artmed_2018_08_001 crossref_primary_10_1016_j_est_2024_114928 crossref_primary_10_1007_s11263_025_02467_7 crossref_primary_10_1109_TIP_2016_2553446 crossref_primary_10_1109_TIP_2018_2886807 crossref_primary_10_1080_09500340_2017_1380854 crossref_primary_10_1109_TMM_2015_2477042 crossref_primary_10_1109_TNNLS_2021_3073248 crossref_primary_10_1145_3644076 crossref_primary_10_1002_mp_13966 crossref_primary_10_1109_TMI_2024_3353762 crossref_primary_10_1109_TPAMI_2020_3046486 crossref_primary_10_1016_j_asoc_2018_10_035 crossref_primary_10_1109_TCSVT_2025_3546031 crossref_primary_10_1007_s11045_017_0505_9 crossref_primary_10_1016_j_patcog_2021_108467 crossref_primary_10_1109_TPAMI_2024_3382138 crossref_primary_10_1371_journal_pone_0295678 crossref_primary_10_1016_j_sigpro_2018_01_013 crossref_primary_10_1016_j_patrec_2025_04_039 crossref_primary_10_3390_math10203736 crossref_primary_10_1109_TPAMI_2016_2578323 crossref_primary_10_1109_TGRS_2020_3023112 crossref_primary_10_1016_j_imavis_2017_09_004 crossref_primary_10_1109_TNNLS_2018_2836969 crossref_primary_10_1016_j_compeleceng_2024_109156 crossref_primary_10_1109_ACCESS_2025_3548510 crossref_primary_10_1080_00036846_2021_1958141 crossref_primary_10_1145_3643675 crossref_primary_10_1049_iet_cvi_2018_5096 crossref_primary_10_1007_s00138_018_0928_9 crossref_primary_10_1093_biomet_asaa011 crossref_primary_10_1109_TCYB_2024_3357213 crossref_primary_10_1007_s40747_024_01360_z crossref_primary_10_1016_j_eswa_2024_125551 crossref_primary_10_1016_j_watres_2023_120773 crossref_primary_10_1109_JSEN_2021_3140081 crossref_primary_10_1109_TNNLS_2016_2551724 crossref_primary_10_1109_TCYB_2016_2572306 crossref_primary_10_1016_j_sigpro_2017_01_021 crossref_primary_10_1016_j_neucom_2016_08_018 crossref_primary_10_1109_TIP_2019_2893066 crossref_primary_10_1109_TMM_2021_3056879 crossref_primary_10_1016_j_ins_2023_119852 crossref_primary_10_1016_j_patcog_2024_110902 crossref_primary_10_1007_s11432_015_5424_5 crossref_primary_10_1109_TCYB_2017_2774266 crossref_primary_10_1109_TKDE_2018_2873791 crossref_primary_10_1007_s10618_018_0603_9 crossref_primary_10_1007_s10994_023_06478_8 crossref_primary_10_1109_TCYB_2021_3070005 crossref_primary_10_3390_w17030401 crossref_primary_10_1007_s11263_024_02064_0 crossref_primary_10_1109_TKDE_2025_3530916 crossref_primary_10_1080_01431161_2024_2339201 crossref_primary_10_1109_TNNLS_2023_3347633 crossref_primary_10_1109_TNSE_2024_3483295 crossref_primary_10_1016_j_ress_2025_111369 crossref_primary_10_1109_JSTARS_2019_2954865 crossref_primary_10_1016_j_procs_2023_08_049 crossref_primary_10_1109_TMM_2021_3116417 crossref_primary_10_1109_TNNLS_2022_3202989 crossref_primary_10_1007_s10044_023_01180_w crossref_primary_10_1007_s11063_016_9550_x crossref_primary_10_1109_TAFFC_2021_3131621 crossref_primary_10_1109_TPAMI_2020_3013379 crossref_primary_10_1103_PhysRevApplied_14_064026 crossref_primary_10_1016_j_patcog_2022_109080 crossref_primary_10_1109_TPAMI_2024_3466182 crossref_primary_10_1109_TCYB_2019_2891577 crossref_primary_10_1109_TBDATA_2016_2515640 crossref_primary_10_1109_TKDE_2021_3061215 crossref_primary_10_1016_j_knosys_2020_106652 crossref_primary_10_1109_TIP_2016_2615423 crossref_primary_10_1109_TMM_2016_2597007 crossref_primary_10_1109_TPAMI_2023_3296156 crossref_primary_10_1109_TKDE_2023_3312109 crossref_primary_10_1109_TPAMI_2024_3388209 crossref_primary_10_1038_s41598_023_43864_7 crossref_primary_10_1109_TIP_2016_2612829 crossref_primary_10_3390_rs13091689 crossref_primary_10_1109_ACCESS_2020_2974779 crossref_primary_10_1007_s00521_020_05343_2 crossref_primary_10_1093_jamia_ocz170 crossref_primary_10_1109_TPAMI_2021_3094362 crossref_primary_10_1109_TPAMI_2023_3236459 crossref_primary_10_1016_j_knosys_2024_112101 crossref_primary_10_1016_j_knosys_2021_107234 crossref_primary_10_1109_TPAMI_2019_2941684 crossref_primary_10_1016_j_cmpb_2021_106327 crossref_primary_10_1016_j_knosys_2017_09_003 crossref_primary_10_1109_TMI_2023_3279110 crossref_primary_10_1109_TIP_2018_2812083 crossref_primary_10_1155_2017_1735176 crossref_primary_10_1109_TIP_2016_2612827 crossref_primary_10_3390_rs11091116 crossref_primary_10_1109_TIP_2016_2593344 crossref_primary_10_1109_TKDE_2023_3279139 crossref_primary_10_3390_make7030095 crossref_primary_10_1007_s11042_017_5172_1 crossref_primary_10_1002_mp_15799 crossref_primary_10_1007_s10994_025_06841_x crossref_primary_10_1109_ACCESS_2021_3088901 crossref_primary_10_1016_j_autcon_2018_11_013 crossref_primary_10_1109_TNNLS_2021_3131377 crossref_primary_10_1007_s13369_020_04869_w crossref_primary_10_1016_j_ijar_2021_04_002 crossref_primary_10_1109_TCYB_2016_2521428 crossref_primary_10_1016_j_neucom_2025_130985 crossref_primary_10_1016_j_sigpro_2017_05_030 crossref_primary_10_1145_3641285 crossref_primary_10_1016_j_eswa_2023_120013 crossref_primary_10_1007_s11517_024_03035_w crossref_primary_10_1016_j_knosys_2021_107907 crossref_primary_10_1109_TPAMI_2024_3387349 |
| Cites_doi | 10.1006/jcom.2002.0642 10.1145/1968.1972 10.1587/transfun.E93.A.787 10.1214/12-EJS699 10.1007/s10107-004-0552-5 10.1007/978-1-4757-3264-1 10.1198/016214505000000907 10.1145/324133.324221 10.1007/BF00116829 10.1109/TIT.2011.2164053 10.1016/S0304-3975(01)00403-0 10.1109/TPAMI.2010.220 10.1111/j.1467-9868.2005.00503.x 10.1214/009053605000000282 10.1007/978-3-540-75225-7_27 10.1109/TPAMI.2008.225 10.1109/TSMCB.2012.2223460 10.1109/SFCS.1997.646140 10.1145/180139.181176 10.1016/0020-0190(96)00006-3 10.1007/PL00013833 10.1137/090763184 10.1109/TPAMI.2009.57 10.1162/NECO_a_00442 10.1145/293347.293351 10.1109/TSP.2007.896065 10.1137/0222052 10.1109/TNNLS.2013.2292894 |
| ContentType | Journal Article |
| Copyright | Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2016 |
| Copyright_xml | – notice: Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2016 |
| DBID | 97E RIA RIE AAYXX CITATION NPM 7SC 7SP 8FD JQ2 L7M L~C L~D 7X8 |
| DOI | 10.1109/TPAMI.2015.2456899 |
| DatabaseName | IEEE Xplore (IEEE) IEEE All-Society Periodicals Package (ASPP) 1998–Present IEEE Electronic Library (IEL) CrossRef PubMed Computer and Information Systems Abstracts Electronics & Communications Abstracts Technology Research Database ProQuest Computer Science Collection Advanced Technologies Database with Aerospace Computer and Information Systems Abstracts Academic Computer and Information Systems Abstracts Professional MEDLINE - Academic |
| DatabaseTitle | CrossRef PubMed Technology Research Database Computer and Information Systems Abstracts – Academic Electronics & Communications Abstracts ProQuest Computer Science Collection Computer and Information Systems Abstracts Advanced Technologies Database with Aerospace Computer and Information Systems Abstracts Professional MEDLINE - Academic |
| DatabaseTitleList | MEDLINE - Academic PubMed Technology Research Database |
| Database_xml | – sequence: 1 dbid: NPM name: PubMed url: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed sourceTypes: Index Database – sequence: 2 dbid: RIE name: IEEE Electronic Library (IEL) url: https://ieeexplore.ieee.org/ sourceTypes: Publisher – sequence: 3 dbid: 7X8 name: MEDLINE - Academic url: https://search.proquest.com/medline sourceTypes: Aggregation Database |
| DeliveryMethod | fulltext_linktorsrc |
| Discipline | Engineering Computer Science |
| EISSN | 2160-9292 1939-3539 |
| EndPage | 461 |
| ExternalDocumentID | 4010178461 27046490 10_1109_TPAMI_2015_2456899 7159100 |
| Genre | orig-research Research Support, Non-U.S. Gov't Journal Article |
| GrantInformation_xml | – fundername: Australian Research Council Projects grantid: DP-140102164; FT-130101457 funderid: 10.13039/501100000923 |
| GroupedDBID | --- -DZ -~X .DC 0R~ 29I 4.4 53G 5GY 6IK 97E AAJGR AARMG AASAJ AAWTH ABAZT ABQJQ ABVLG ACGFO ACGFS ACIWK ACNCT AENEX AGQYO AGSQL AHBIQ AKQYR ALMA_UNASSIGNED_HOLDINGS ASUFR ATWAV BEFXN BFFAM BGNUA BKEBE BPEOZ CS3 DU5 E.L EBS EJD F5P HZ~ IEDLZ IFIPE IPLJI JAVBF LAI M43 MS~ O9- OCL P2P PQQKQ RIA RIE RNS RXW TAE TN5 UHB ~02 AAYXX CITATION 5VS 9M8 ABFSI ADRHT AETEA AETIX AI. AIBXA AKJIK ALLEH FA8 H~9 IBMZZ ICLAB IFJZH NPM RIG RNI RZB VH1 XJT 7SC 7SP 8FD JQ2 L7M L~C L~D 7X8 |
| ID | FETCH-LOGICAL-c351t-c29fc4e3fa98693b4f52b8da3f07ba36dafe9ac87eedc66c3bc4b1fbfe0dd20f3 |
| IEDL.DBID | RIE |
| ISICitedReferencesCount | 673 |
| ISICitedReferencesURI | http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=000370738900003&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D |
| ISSN | 0162-8828 1939-3539 |
| IngestDate | Wed Oct 01 14:45:54 EDT 2025 Sun Nov 09 07:38:33 EST 2025 Mon Jul 21 06:01:45 EDT 2025 Sat Nov 29 05:15:56 EST 2025 Tue Nov 18 21:58:06 EST 2025 Wed Aug 27 02:22:29 EDT 2025 |
| IsPeerReviewed | true |
| IsScholarly | true |
| Issue | 3 |
| Keywords | label noise noise rate estimation importance reweighting consistency Classification |
| Language | English |
| License | https://ieeexplore.ieee.org/Xplorehelp/downloads/license-information/IEEE.html https://doi.org/10.15223/policy-029 https://doi.org/10.15223/policy-037 |
| LinkModel | DirectLink |
| MergedId | FETCHMERGED-LOGICAL-c351t-c29fc4e3fa98693b4f52b8da3f07ba36dafe9ac87eedc66c3bc4b1fbfe0dd20f3 |
| Notes | ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 14 content type line 23 |
| PMID | 27046490 |
| PQID | 1778425651 |
| PQPubID | 85458 |
| PageCount | 15 |
| ParticipantIDs | crossref_primary_10_1109_TPAMI_2015_2456899 proquest_journals_1778425651 pubmed_primary_27046490 crossref_citationtrail_10_1109_TPAMI_2015_2456899 proquest_miscellaneous_1779414563 ieee_primary_7159100 |
| PublicationCentury | 2000 |
| PublicationDate | 2016-March-1 2016-3-1 2016-Mar 20160301 |
| PublicationDateYYYYMMDD | 2016-03-01 |
| PublicationDate_xml | – month: 03 year: 2016 text: 2016-March-1 day: 01 |
| PublicationDecade | 2010 |
| PublicationPlace | United States |
| PublicationPlace_xml | – name: United States – name: New York |
| PublicationTitle | IEEE transactions on pattern analysis and machine intelligence |
| PublicationTitleAbbrev | TPAMI |
| PublicationTitleAlternate | IEEE Trans Pattern Anal Mach Intell |
| PublicationYear | 2016 |
| Publisher | IEEE The Institute of Electrical and Electronics Engineers, Inc. (IEEE) |
| Publisher_xml | – name: IEEE – name: The Institute of Electrical and Electronics Engineers, Inc. (IEEE) |
| References | ref12 ref15 gretton (ref39) 2009 ref14 ref11 moore (ref29) 0 belkin (ref13) 2006; 7 ref16 ref19 ref18 yang (ref6) 0 ref51 biggio (ref5) 0 vapnik (ref45) 0 klivans (ref22) 2009; 10 lawrence (ref4) 0 ref47 ref42 he (ref31) 2011; 33 bartlett (ref41) 2003; 3 ref44 ref43 anthony (ref37) 2009 ref8 crammer (ref32) 0 ref7 long (ref21) 0 ref9 ref3 scott (ref33) 0 ref36 sun (ref50) 0 mohri (ref40) 2012 ref30 huang (ref48) 0 sugiyama (ref46) 2010 ref2 ref1 ref38 natarajan (ref10) 0 sugiyama (ref52) 0 ref24 ref23 ref26 khardon (ref28) 2007; 8 ref25 kanamori (ref49) 2009; 10 ref20 ref27 blanchard (ref34) 2010; 11 gong (ref17) 0 scott (ref35) 0 |
| References_xml | – year: 0 ident: ref29 publication-title: Robust Regression Using Maximum-Likelihood Weighting and Assuming Cauchy-Distributed Random Error – volume: 11 start-page: 2973 year: 2010 ident: ref34 article-title: Semi-supervised novelty detection publication-title: J Mach Learn Res – start-page: 1433 year: 0 ident: ref52 article-title: Direct importance estimation with model selection and its application to covariate shift adaptation publication-title: Proc Adv Neural Inf Process Syst 20 – ident: ref43 doi: 10.1006/jcom.2002.0642 – start-page: 306 year: 0 ident: ref4 article-title: Estimating a kernel Fisher discriminant in the presence of label noise publication-title: Proc 8th Int Conf Mach Learn – ident: ref2 doi: 10.1145/1968.1972 – volume: 10 start-page: 2715 year: 2009 ident: ref22 article-title: Learning halfspaces with malicious noise publication-title: J Mach Learn Res – ident: ref47 doi: 10.1587/transfun.E93.A.787 – ident: ref12 doi: 10.1214/12-EJS699 – ident: ref16 doi: 10.1007/s10107-004-0552-5 – start-page: 97 year: 0 ident: ref5 article-title: Support vector machines under adversarial label noise publication-title: Proc Asian Conf Mach Learn – ident: ref36 doi: 10.1007/978-1-4757-3264-1 – volume: 8 start-page: 227 year: 2007 ident: ref28 article-title: Noise tolerant variants of the perceptron algorithm publication-title: J Mach Learn Res – ident: ref11 doi: 10.1198/016214505000000907 – ident: ref19 doi: 10.1145/324133.324221 – volume: 3 start-page: 463 year: 2003 ident: ref41 article-title: Rademacher and Gaussian complexities: Risk bounds and structural results publication-title: J Mach Learn Res – ident: ref1 doi: 10.1007/BF00116829 – ident: ref23 doi: 10.1109/TIT.2011.2164053 – ident: ref20 doi: 10.1016/S0304-3975(01)00403-0 – start-page: 838 year: 0 ident: ref35 article-title: A rate of convergence for mixture proportion estimation, with application to learning from noisy labels publication-title: Proc 18th Int Conf Artif Intell Statist – volume: 33 start-page: 1561 year: 2011 ident: ref31 article-title: Maximum correntropy criterion for robust face recognition publication-title: IEEE Trans Pattern Anal Mach Intell doi: 10.1109/TPAMI.2010.220 – start-page: 91 year: 0 ident: ref21 article-title: Learning large-margin halfspaces with more malicious noise publication-title: Proc 25th Annu Conf Neural Inf Process Syst – ident: ref14 doi: 10.1111/j.1467-9868.2005.00503.x – ident: ref42 doi: 10.1214/009053605000000282 – start-page: 1196 year: 0 ident: ref10 article-title: Learning with noisy labels publication-title: Proc Neural Inf Process Syst – ident: ref27 doi: 10.1007/978-3-540-75225-7_27 – start-page: 37 year: 0 ident: ref17 article-title: A general iterative shrinkage and thresholding algorithm for non-convex regularized optimization problems publication-title: Proc Int Conf Mach Learn – start-page: 601 year: 0 ident: ref48 article-title: Correcting sample selection bias by unlabeled data publication-title: Proc Adv Neural Inf Process Syst 19 – start-page: 489 year: 0 ident: ref33 article-title: Classification with asymmetric label noise: Consistency and maximal denoising publication-title: Proc Conf Learn Theory – ident: ref51 doi: 10.1109/TPAMI.2008.225 – ident: ref9 doi: 10.1109/TSMCB.2012.2223460 – year: 2012 ident: ref40 publication-title: Foundations of Machine Learning – ident: ref25 doi: 10.1109/SFCS.1997.646140 – start-page: 10 year: 2010 ident: ref46 article-title: Density ratio estimation: A comprehensive review publication-title: RIMS Kokyuroku – volume: 7 start-page: 2399 year: 2006 ident: ref13 article-title: Manifold regularization: A geometric framework for learning from labeled and unlabeled examples publication-title: J Mach Learn Res – year: 0 ident: ref45 article-title: Constructive setting of the density ratio estimation problem and its rigorous solution publication-title: arXiv preprint arXiv 1306 0407 – ident: ref24 doi: 10.1145/180139.181176 – year: 2009 ident: ref37 publication-title: Neural Network Learning Theoretical Foundations – volume: 10 start-page: 1391 year: 2009 ident: ref49 article-title: A least-squares approach to direct importance estimation publication-title: J Mach Learn Res – ident: ref8 doi: 10.1016/0020-0190(96)00006-3 – start-page: 505 year: 0 ident: ref50 article-title: A two-stage weighting framework for multi-source domain adaptation publication-title: Proc Adv Neural Inf Process Syst 24 – ident: ref26 doi: 10.1007/PL00013833 – ident: ref15 doi: 10.1137/090763184 – ident: ref38 doi: 10.1109/TPAMI.2009.57 – ident: ref44 doi: 10.1162/NECO_a_00442 – ident: ref3 doi: 10.1145/293347.293351 – start-page: 131 year: 2009 ident: ref39 article-title: Covariate shift by kernel mean matching publication-title: Dataset Shift in Machine Learning – start-page: 451 year: 0 ident: ref32 article-title: Learning via Gaussian herding publication-title: Proc Adv Neural Inf Process Syst 23 – ident: ref30 doi: 10.1109/TSP.2007.896065 – ident: ref18 doi: 10.1137/0222052 – start-page: 233 year: 0 ident: ref6 article-title: Multiple kernel learning from noisy labels by stochastic programming publication-title: Proc 29th Int Conf Mach Learn – ident: ref7 doi: 10.1109/TNNLS.2013.2292894 |
| SSID | ssj0014503 |
| Score | 2.665921 |
| Snippet | In this paper, we study a classification problem in which sample labels are randomly corrupted. In this scenario, there is an unobservable sample with... |
| SourceID | proquest pubmed crossref ieee |
| SourceType | Aggregation Database Index Database Enrichment Source Publisher |
| StartPage | 447 |
| SubjectTerms | Algorithm design and analysis Classification consistency Convergence Estimation importance reweighting Kernel label noise Noise Noise measurement noise rate estimation Robustness |
| Title | Classification with Noisy Labels by Importance Reweighting |
| URI | https://ieeexplore.ieee.org/document/7159100 https://www.ncbi.nlm.nih.gov/pubmed/27046490 https://www.proquest.com/docview/1778425651 https://www.proquest.com/docview/1779414563 |
| Volume | 38 |
| WOSCitedRecordID | wos000370738900003&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D |
| hasFullText | 1 |
| inHoldings | 1 |
| isFullTextHit | |
| isPrint | |
| journalDatabaseRights | – providerCode: PRVIEE databaseName: IEEE Electronic Library (IEL) customDbUrl: eissn: 2160-9292 dateEnd: 99991231 omitProxy: false ssIdentifier: ssj0014503 issn: 0162-8828 databaseCode: RIE dateStart: 19790101 isFulltext: true titleUrlDefault: https://ieeexplore.ieee.org/ providerName: IEEE |
| link | http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV3dS-QwEB9U7uF8OE-9j6onEXzzqu2mTRrf5LhF4VxEFPat5GMCgu4e7u4d-987ST_wQQXfCk3akpnp_Ca_yQzAoVGVLiQPrC4WaaHQpFWlypRro5AQqVOZi80m5GhUjcfqagV-9mdhEDEmn-FxuIxcvpvaRdgqO5Hke_OMAvRVKWVzVqtnDIoydkEmBEMWTmFEd0AmUyc3V2eXFyGLqzwONB9FGKEEsAykXvgVP_NHscHK61gz-pzhxvu-9jN8arElO2uUYRNWcLIFG13fBtaa8RasPytCuA2nsS9myBiKQmJhZ5aNpnezJfujDblOZpbs4iHidJrOrvF_3E6lyV_gdvj75td52nZUSC0v83lqB8rbArnXqhKKm8KXA1M5zX0mjebCaY9K20qS57RCWG5sYXJvPGbODTLPv8LaZDrB78AcAR9CP8Z5pBCTRE4wHSmeFl66zAibQN6ta23bcuOh68V9HcOOTNVRLHUQS92KJYGjfs7fptjGm6O3w6L3I9v1TmCvE1_d2uOszqUMfKMo8wQO-ttkSYEe0ROcLuIYVZAeCZ7At0bs_bM7bdl5-Z278JG-TDS5aXuwNn9c4A_4YP_N72aP-6Su42o_qusT6RTj3g |
| linkProvider | IEEE |
| linkToHtml | http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV3daxQxEB9KK2gfWm1t3Vo1gm-67e4lm2x8K2Lp4fUockLflnxMoKB30rtT-t93kv2gDyr4trDJ7pKZ2flNfpMZgHdW10YoHlldFLnQaPO61lXOjdVIiNTrwqdmE2o6ra-v9dUGfBjOwiBiSj7Dk3iZuHy_cOu4VXaqyPeWBQXoW5UQo7I9rTVwBqJKfZAJw5CNUyDRH5Ep9Ons6uxyHPO4qpNI9FGMEYsAq0jrxZ_xA4-UWqz8HW0mr3O--3_f-xR2OnTJzlp1eAYbON-D3b5zA-sMeQ-2H5Qh3IePqTNmzBlKYmJxb5ZNFzfLOzYxlpwns3ds_CMhdZrOvuLvtKFKk5_Dt_PPs08XeddTIXe8Kle5G-ngBPJgdC01tyJUI1t7w0OhrOHSm4DauFqR73RSOm6dsGWwAQvvR0XgB7A5X8zxBTBP0Ifwj_UBKcgkoRNQR4qoZVC-sNJlUPbr2riu4Hjse_G9SYFHoZskliaKpenEksH7Yc7PttzGP0fvx0UfRnbrncFxL76ms8hlUyoVGUdZlRm8HW6TLUWCxMxxsU5jtCA9kjyDw1bsw7N7bTn68zvfwOOL2eWkmYynX17CE_pK2WaqHcPm6naNr-CR-7W6Wd6-Tkp7D4qT5j0 |
| openUrl | ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=Classification+with+Noisy+Labels+by+Importance+Reweighting&rft.jtitle=IEEE+transactions+on+pattern+analysis+and+machine+intelligence&rft.au=Liu%2C+Tongliang&rft.au=Tao%2C+Dacheng&rft.date=2016-03-01&rft.issn=0162-8828&rft.eissn=2160-9292&rft.volume=38&rft.issue=3&rft.spage=447&rft.epage=461&rft_id=info:doi/10.1109%2FTPAMI.2015.2456899&rft.externalDBID=n%2Fa&rft.externalDocID=10_1109_TPAMI_2015_2456899 |
| thumbnail_l | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=0162-8828&client=summon |
| thumbnail_m | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=0162-8828&client=summon |
| thumbnail_s | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=0162-8828&client=summon |