Analysis of Learning from Positive and Unlabeled Data

By Marthinus Christoffel du Plessis, Gang Niu, Masashi Sugiyama. Neural Information Processing Systems, 2014

438 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/03b3f0b58e14d7840303e52b17e023fb54667fdc

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

On Symmetric Losses for Learning from Corrupted Labels (2019, background): “…that satisﬁes a symmetric condition has demonstrated its usefulness in weakly-supervised learning, e.g., one can use a symmetric loss to simplify a risk estimator in learning from positive-unlabeled data [du Plessis et al., 2014] and learning from two sets of unlabeled data [Lu et al., 2018].”
Unsupervised Representation Learning for Time Series with Temporal Neighborhood Coding (2021, methodology): “In PU learning, a classiﬁer is learned using labeled data drawn from the positive class ( P ) and unlabeled data ( U ) that is a mixture of positive and negative samples with a positive class prior π (Du Plessis et al., 2014; Kiryo et al., 2017; Du Plessis & Sugiyama, 2014).”
Combating Noisy Labels by Agreement: A Joint Training Method with Co-Regularization (2020, background): “have been developed for learning with noisy labels, which belongs to the family of weakly supervised learning frameworks [2, 5, 6, 7, 8, 9, 11].”
Positive-Unlabeled Learning with Non-Negative Risk Estimator (2017, methodology): “…risk estimators, if we unrealistically assume that the class-posterior probability is one for all P data.1 A breakthrough in this direction is du Plessis et al. (2014), for proposing the first unbiased risk estimator, and a more general estimator was suggested in du Plessis et al. (2015) as a…”
A Unifying Review of Deep and Shallow Anomaly Detection (2020, background): “This is one of the reasons why the special semisupervised case of LPUE [234]–[236], that is, labeled normal and unlabeled examples, is also studied specifically in the AD literature [148], [161], [237]–[239].”
Debiased Contrastive Learning (2020, methodology): “Our approach is related to unbiased PU learning, where the unlabeled data is used as negative examples, but down-weighted appropriately [26, 12, 11].”
Positive-Unlabeled Reward Learning (2019, methodology): “In both cases, training the discriminator is posed as a positive-unlabeled (PU) learning problem (Denis, 1998; Elkan & Noto, 2008; du Plessis et al., 2014), and we are able to train a reliable discriminator with the desired semantics by applying a recent large scale PU learning algorithm (Kiryo et…”
On Positive-Unlabeled Classification in GAN (2020, background): “Identifying high-quality samples from generated samples under the guidance of real samples is very similar to Positive-Unlabeled classification problems [3, 12, 23], where only some positive samples were labeled, and the classifier tried to find positive samples from unlabeled samples consisting of ”
Fast Threshold Tests for Detecting Discrimination (2017, background): “The setup thus bears resemblance to learning classiﬁers from only positive examples (Elkan & Noto, 2008; Mordelet & Vert, 2014; du Plessis et al., 2014).”
Convex Formulation for Learning from Positive and Unlabeled Data (2015, methodology): “• Ramp: The method of du Plessis et al. (2014) using the non-convex ramp loss.”
Robust Loss Functions under Label Noise for Deep Neural Networks (2017, background): “The problem of learning from positive and unlabeled data can be cast as learning under label noise and it is seen that none of the common convex surrogate losses are good for this problem [8].”
Learning from positive and unlabeled data: a survey (2018, methodology): “The ShiftMC method for matrix completion is also a special case of this method under the SCAR assumption, using the MSE loss [36]. du Plessis et al. (2014) proposed another risk estimator, which simply reweights the examples and does not introduce duplicates [80].”
Contrastive Learning with Hard Negative Samples (2020, methodology): “To work towards a practical method, note that we can rewrite this distribution by adopting a PU-learning viewpoint (Elkan & Noto, 2008; Du Plessis et al., 2014; Chuang et al., 2020).”