Ordinal Label Proportions
Abstract
In Machine Learning, it is common to distinguish different degrees of supervision, ranging from fully supervised to completely unsupervised scenarios. However, lying in between those, the Learning from Label Proportions (LLP) setting [ 19 ] assumes the training data is provided in the form of bags, and the only supervision comes through the proportion of each class in each bag. In this paper, we present a novel version of the LLP paradigm where the relationship among the classes is ordinal. While this is a highly relevant scenario (e.g. customer surveys where the results can be divided into various degrees of satisfaction), it is as yet unexplored in the literature. We refer to this setting as Ordinal Label Proportions (OLP). We formally define the scenario and introduce an efficient algorithm to tackle it. We test our algorithm on synthetic and benchmark datasets. Additionally, we present a case study examining a dataset gathered from the Research Excellence Framework that assesses the quality of research in the United Kingdom.
Cite
Text
Poyiadzi et al. "Ordinal Label Proportions." European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, 2018. doi:10.1007/978-3-030-10925-7_19Markdown
[Poyiadzi et al. "Ordinal Label Proportions." European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, 2018.](https://mlanthology.org/ecmlpkdd/2018/poyiadzi2018ecmlpkdd-ordinal/) doi:10.1007/978-3-030-10925-7_19BibTeX
@inproceedings{poyiadzi2018ecmlpkdd-ordinal,
title = {{Ordinal Label Proportions}},
author = {Poyiadzi, Rafael and Santos-Rodríguez, Raúl and De Bie, Tijl},
booktitle = {European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases},
year = {2018},
pages = {306-321},
doi = {10.1007/978-3-030-10925-7_19},
url = {https://mlanthology.org/ecmlpkdd/2018/poyiadzi2018ecmlpkdd-ordinal/}
}