Filters








39,264 Hits in 4.5 sec

Iterative Learning with Open-set Noisy Labels [article]

Yisen Wang, Weiyang Liu, Xingjun Ma, James Bailey, Hongyuan Zha, Le Song, Shu-Tao Xia
2018 arXiv   pre-print
To address this problem, we propose a novel iterative learning framework for training CNNs on datasets with open-set noisy labels.  ...  Our approach detects noisy labels and learns deep discriminative features in an iterative fashion.  ...  In this paper, we propose an iterative learning framework that can robustly train CNNs on datasets with open-set noisy labels.  ... 
arXiv:1804.00092v1 fatcat:56cvrf2yujftllyqs6dcocjste

Iterative Learning with Open-set Noisy Labels

Yisen Wang, Weiyang Liu, Xingjun Ma, James Bailey, Hongyuan Zha, Le Song, Shu-Tao Xia
2018 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition  
To address this problem, we propose a novel iterative learning framework for training CNNs on datasets with open-set noisy labels.  ...  Our approach detects noisy labels and learns deep discriminative features in an iterative fashion.  ...  In this paper, we propose an iterative learning framework that can robustly train CNNs on datasets with open-set noisy labels.  ... 
doi:10.1109/cvpr.2018.00906 dblp:conf/cvpr/WangLMBZSX18 fatcat:ys5xxa6jpvh2fi7evpwqvc77v4

Deep Learning Classification With Noisy Labels [article]

Guillaume Sanchez, Vincente Guis, Ricard Marxer, Frédéric Bouchara
2020 arXiv   pre-print
Deep Learning systems have shown tremendous accuracy in image classification, at the cost of big image datasets. Collecting such amounts of data can lead to labelling errors in the training set.  ...  We review recent works on how to manage noisy annotations when training deep learning classifiers, independently from our interest in face recognition.  ...  open-set and closed-set noise.  ... 
arXiv:2004.11116v1 fatcat:zp4cuhuxpbgxxihgecvqtrlx2q

Deep Learning Classification with Noisy Labels

Guillaume SANCHEZ, Vincente GUIS, Ricard MARXER, Frederic BOUCHARA
2020 2020 IEEE International Conference on Multimedia & Expo Workshops (ICMEW)  
open-set and closed-set noise.  ...  Finally, Iterative Noise Filtering [15] , assumes that the class with the highest prediction for the unlabeled examples is correct. Deep Self-Learning [20] learns an initial net on noisy labels.  ... 
doi:10.1109/icmew46912.2020.9105992 dblp:conf/icmcs/SanchezGMB20 fatcat:f4bhvsqu35db3mn4btbbcjabiu

Attention-Aware Noisy Label Learning for Image Classification [article]

Zhenzhen Wang, Chunyan Xu, Yap-Peng Tan, Junsong Yuan
2020 arXiv   pre-print
In this paper, the attention-aware noisy label learning approach (A^2NL) is proposed to improve the discriminative capability of the network trained on datasets with potential label noise.  ...  The cheapest way to obtain a large body of labeled visual data is to crawl from websites with user-supplied labels, such as Flickr.  ...  The goal of the proposed method does not end with predicting the noisy labelsỹ or learning the noisy distributions, instead it is to predict the true labels given the image sets with noisy labels.  ... 
arXiv:2009.14757v1 fatcat:7f7ympduavhlhn2xqluyfwquxy

EvidentialMix: Learning with Combined Open-set and Closed-set Noisy Labels [article]

Ragav Sachdeva, Filipe R. Cordeiro, Vasileios Belagiannis, Ian Reid, Gustavo Carneiro
2020 arXiv   pre-print
In this work, we study a new variant of the noisy label problem that combines the open-set and closed-set noisy labels, and introduce a benchmark evaluation to assess the performance of training algorithms  ...  However, acquiring such large-scale data sets with precise annotations is very expensive and time-consuming, and the cheap alternatives often yield data sets that have noisy labels.  ...  The learning with open-set noisy labels has only recently been explored by Wang et al.  ... 
arXiv:2011.05704v1 fatcat:7y2yn3fpcfarvfoondzzpq4y2y

Noise-resistant Deep Metric Learning with Ranking-based Instance Selection [article]

Chang Liu and Han Yu and Boyang Li and Zhiqi Shen and Zhanning Gao and Peiran Ren and Xuansong Xie and Lizhen Cui and Chunyan Miao
2021 arXiv   pre-print
Although much research effort has been devoted to improving robustness to noisy labels in classification tasks, the problem of noisy labels in deep metric learning (DML) remains open.  ...  The existence of noisy labels in real-world data negatively impacts the performance of deep learning models.  ...  In classification, [47] simulates open-set noisy labels by adding data from other datasets. In this paper, we propose the Small Cluster noise model for open-set label noise in metric learning.  ... 
arXiv:2103.16047v2 fatcat:es66xlztzbc4rovic432bzxtla

S3: Supervised Self-supervised Learning under Label Noise [article]

Chen Feng, Georgios Tzimiropoulos, Ioannis Patras
2021 arXiv   pre-print
In this context, in this paper we address the problem of classification in the presence of label noise and more specifically, both close-set and open-set label noise, that is when the true label of a sample  ...  Despite the large progress in supervised learning with Neural Networks, there are significant challenges in obtaining high-quality, large-scale and accurately labeled datasets.  ...  Clearly, for an open-set noisy label it is the case that y i = y i , y i / ∈ {0, 1} K , while for closed-set noisy samples y i = y i , y i ∈ {0, 1} K .  ... 
arXiv:2111.11288v1 fatcat:lqudfwtx5zcffelsuxvhew4wwq

Automatic Error Correction for Speaker Embedding Learning with Noisy Labels

Fuchuan Tong, Yan Liu, Song Li, Jie Wang, Lin Li, Qingyang Hong
2021 Conference of the International Speech Communication Association  
However, noisy labels often occur during data collection. In this paper, we propose an automatic error correction method for deep speaker embedding learning with noisy labels.  ...  Moreover, when combining with the Bayesian estimation of PLDA with noisy training labels at the back-end, the whole system performs better under conditions in which noisy labels are present.  ...  Related works Network learning with noisy labels Types of label noise can be categorized into two categories: closed-set noise and open-set noise.  ... 
doi:10.21437/interspeech.2021-2021 dblp:conf/interspeech/TongLLWLH21 fatcat:6ynqdzftlrckhbzuvo3fbu42ey

Trade When Opportunity Comes: Price Movement Forecasting via Locality-Aware Attention and Iterative Refinement Labeling [article]

Liang Zeng, Lei Wang, Hui Niu, Jian Li, Ruchen Zhang, Zhonghao Dai, Dewei Zhu, Ling Wang
2022 arXiv   pre-print
)Iterative refinement labeling further iteratively refines the labels of noisy samples and then combines the learned predictors to be robust to the unseen and noisy samples.  ...  set of machine learning based competitors on the Qlib platform.  ...  Thus, it inspires us to further denoise the labels of noisy samples with multiple iterative evaluations.  ... 
arXiv:2107.11972v2 fatcat:or3ymqiljjhyrk5zc74ll37bme

Label Refinement with an Iterative Generative Adversarial Network for Boosting Retinal Vessel Segmentation [article]

Yunqiao Yang, Zhiwei Wang, Jingen Liu, Kwang-Ting Cheng, Xin Yang
2019 arXiv   pre-print
Our iterative GAN is trained based on a set of high-quality patches (i.e. with consistent manual labels among different observers) and low-quality patches with noisy manual vessel labels.  ...  To this end, we have developed a novel label refinement method based on an iterative generative adversarial network (GAN).  ...  Our training data is collected based on a set of high-quality retinal patches with consistent manual labels among different observers and low-quality retinal patches with noisy manual vessel labels.  ... 
arXiv:1912.02589v1 fatcat:ffewyn434bhxlfdjeyvmlu2xhm

Sample Noise Impact on Active Learning [article]

Alexandre Abraham, Léo Dreyfus-Schmidt
2021 arXiv   pre-print
This work explores the effect of noisy sample selection in active learning strategies.  ...  We hope that the questions raised in this paper are of interest to the community and could open new paths for active learning research.  ...  We use KCenterGreedy (KCenter) as a proxy for Core-sets [6] since there is no open implementation available.  ... 
arXiv:2109.01372v1 fatcat:f4jav3ap6bcqznyjeotn3dhlnm

Learning from Noisy Labels with No Change to the Training Process

Mingyuan Zhang, Jane Lee, Shivani Agarwal
2021 International Conference on Machine Learning  
There has been much interest in recent years in developing learning algorithms that can learn accurate classifiers from data with noisy labels.  ...  A widely-studied noise model is that of classconditional noise (CCN), wherein a label y is flipped to a label y with some associated noise probability that depends on both y and y.  ...  The primary challenge in learning from noisy labels is to design algorithms which, despite being given data with noisy labels as input, can learn accurate classifiers for the true, clean distribution.  ... 
dblp:conf/icml/ZhangL021 fatcat:nma4wppejvfcfphxbhmjj57kra

Open-set Label Noise Can Improve Robustness Against Inherent Label Noise [article]

Hongxin Wei, Lue Tao, Renchunzi Xie, Bo An
2021 arXiv   pre-print
Learning with noisy labels is a practically challenging problem in weakly supervised learning.  ...  In this paper, we empirically show that open-set noisy labels can be non-toxic and even benefit the robustness against inherent noisy labels.  ...  It is worthy to note that open-set noisy labels do not encounter this issue. Learning with inherent noisy labels.  ... 
arXiv:2106.10891v3 fatcat:2vjoezev7za4bbmzz5ipqakp5y

How does Disagreement Help Generalization against Label Corruption? [article]

Xingrui Yu, Bo Han, Jiangchao Yao, Gang Niu, Ivor W. Tsang, Masashi Sugiyama
2019 arXiv   pre-print
Learning with noisy labels is one of the hottest problems in weakly-supervised learning.  ...  Based on memorization effects of deep neural networks, training on small-loss instances becomes very promising for handling noisy labels.  ...  We gratefully acknowledge the support of NVIDIA Corporation with the donation of Titan Xp GPU used for this research.  ... 
arXiv:1901.04215v3 fatcat:fugxwrwouja3focmitlgsxze2y
« Previous Showing results 1 — 15 out of 39,264 results