Learning from positive and unlabeled data: a survey

By Jessa Bekker, Jesse Davis. 2018

706 citations recorded by the source corpus.

Learning from positive and unlabeled data or PU learning is the setting where a learner only has access to positive examples and unlabeled data. The assumption is that the unlabeled data can contain both positive and negative examples. This setting has attracted increasing interest within the machine learning literature as this type of data naturally arises in applications such as medical diagnosis and knowledge base completion. This article provides a survey of the current state of the art in PU learning. It proposes seven key research questions that commonly arise in this field and provides a broad overview of how the field has tried to address them.

Subfields: Artificial Intelligence

How researchers see this paper

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

In Artificial Intelligence

1. Dacheng Tao
2. Ziwei Liu
3. Kaiyang Zhou
4. Jingkang Yang
5. Quanming Yao
6. Qiang Yang
7. Wei-Wei Tu
8. Yang Yu

Citations in the corpus

• Out-of-Domain Detection for Natural Language Understanding in Dialog Systems (2019, background): “Another branch of related studies is the Positive and Unlabeled (PU) learning [40], in which a learning algorithm has only access to positive examples and unlabeled data.”
• Automated Machine Learning: From Principles to Practices (2018, background): “For example, Landmark regularization is proposed by [160] to improve the consistency between the performance rankings of the supernet model and that of the standalone architectures.”
• A Survey on Deep Semi-Supervised Learning (2021, background): “Positive and unlabeled (PU) learning [48], [49] is a variant of positive and negative binary classification, where the training data consists of positive samples and unlabeled samples.”
• Self-Supervised Representation Learning by Rotation Feature Decoupling (2019, background): “Predicting whether an input image is rotated is hence a binary classification problem if all unlabeled data are treated as negative examples [4].”
• Generalized Out-of-Distribution Detection: A Survey (2021, positive): “PU learning (Zhang & Zuo, 2008; Bekker & Davis, 2020; Jaskie & Spanias, 2019; Ruff et al., 2020) is proposed for the semi-supervised AD setting where unlabeled data is available in addition to the normal data.”