arXiv subjects

Paul Albert

Publications and source records attributed to Paul Albert.

22 records · Page 2Linked to original sources

arXiv2020

Towards Robust Learning with Different Label Noise Distributions

Noisy labels are an unavoidable consequence of labeling processes and detecting them is an important step towards preventing performance degradations in Convolutional Neural Networks. Discarding noisy labels avoids a harmful memorization, while the associated image content can still be exploited in a semi-supervised learning (SSL) setup. Clean samples are usually identified using the small loss trick, i.e. they exhibit a low loss. However, we show that different noise distributions make the application of this trick less straightforward and propose to continuously relabel all images to reveal a discriminative loss against multiple distributions. SSL is then applied twice, once to improve the clean-noisy detection and again for training the final model. We design an experimental setup based on ImageNet32/64 for better understanding the consequences of representation learning with differing label noise distributions and find that non-uniform out-of-distribution noise better resembles real-world noise and that in most cases intermediate features are not affected by label noise corruption. Experiments in CIFAR-10/100, ImageNet32/64 and WebVision (real-world noise) demonstrate that the proposed label noise Distribution Robust Pseudo-Labeling (DRPL) approach gives substantial improvements over recent state-of-the-art. Code is available at https://git.io/JJ0PV.

cs.CV↗

arXiv2020

Pseudo-Labeling and Confirmation Bias in Deep Semi-Supervised Learning

Semi-supervised learning, i.e. jointly learning from labeled and unlabeled samples, is an active research topic due to its key role on relaxing human supervision. In the context of image classification, recent advances to learn from unlabeled samples are mainly focused on consistency regularization methods that encourage invariant predictions for different perturbations of unlabeled samples. We, conversely, propose to learn from unlabeled data by generating soft pseudo-labels using the network predictions. We show that a naive pseudo-labeling overfits to incorrect pseudo-labels due to the so-called confirmation bias and demonstrate that mixup augmentation and setting a minimum number of labeled samples per mini-batch are effective regularization techniques for reducing it. The proposed approach achieves state-of-the-art results in CIFAR-10/100, SVHN, and Mini-ImageNet despite being much simpler than other methods. These results demonstrate that pseudo-labeling alone can outperform consistency regularization methods, while the opposite was supposed in previous work. Source code is available at https://git.io/fjQsC.

cs.CV↗

arXiv2019

Unsupervised Label Noise Modeling and Loss Correction

Despite being robust to small amounts of label noise, convolutional neural networks trained with stochastic gradient methods have been shown to easily fit random labels. When there are a mixture of correct and mislabelled targets, networks tend to fit the former before the latter. This suggests using a suitable two-component mixture model as an unsupervised generative model of sample loss values during training to allow online estimation of the probability that a sample is mislabelled. Specifically, we propose a beta mixture to estimate this probability and correct the loss by relying on the network prediction (the so-called bootstrapping loss). We further adapt mixup augmentation to drive our approach a step further. Experiments on CIFAR-10/100 and TinyImageNet demonstrate a robustness to label noise that substantially outperforms recent state-of-the-art. Source code is available at https://git.io/fjsvE

cs.CV↗

arXiv2017

Sequential estimation in the group testing problem

Estimation using pooled sampling has long been an area of interest in the group testing literature. Such research has focused primarily on the assumed use of fixed sampling plans (i), although some recent papers have suggested alternative sequential designs that sample until a predetermined number of positive tests (ii). One major consideration, including in the new work on sequential plans, is the construction of debiased estimators which either reduce or keep the mean square error from inflating. Whether, however, under the above or other sampling designs unbiased estimation is in fact possible has yet to be established in the literature. In this paper, we introduce a design which samples until a fixed number of negatives (iii), and show that an unbiased estimator exists under this model, while unbiased estimation is not possible for either of the preceding designs (i) and (ii). We present new estimators under the different sampling plans that are either unbiased or that have reduced bias relative to those already in use as well as generally improve on the mean square error. Numerical studies are done in order to compare designs in terms of bias and mean square error under practical situations with small and medium sample sizes.

math.ST↗