ProPML: Probability Partial Multi-label Learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Struski, Łukasz, Pardyl, Adam, Tabor, Jacek, Zieliński, Bartosz
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929273591300096
author Struski, Łukasz
Pardyl, Adam
Tabor, Jacek
Zieliński, Bartosz
author_facet Struski, Łukasz
Pardyl, Adam
Tabor, Jacek
Zieliński, Bartosz
contents Partial Multi-label Learning (PML) is a type of weakly supervised learning where each training instance corresponds to a set of candidate labels, among which only some are true. In this paper, we introduce \our{}, a novel probabilistic approach to this problem that extends the binary cross entropy to the PML setup. In contrast to existing methods, it does not require suboptimal disambiguation and, as such, can be applied to any deep architecture. Furthermore, experiments conducted on artificial and real-world datasets indicate that \our{} outperforms existing approaches, especially for high noise in a candidate set.
format Preprint
id arxiv_https___arxiv_org_abs_2403_07603
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle ProPML: Probability Partial Multi-label Learning
Struski, Łukasz
Pardyl, Adam
Tabor, Jacek
Zieliński, Bartosz
Machine Learning
Partial Multi-label Learning (PML) is a type of weakly supervised learning where each training instance corresponds to a set of candidate labels, among which only some are true. In this paper, we introduce \our{}, a novel probabilistic approach to this problem that extends the binary cross entropy to the PML setup. In contrast to existing methods, it does not require suboptimal disambiguation and, as such, can be applied to any deep architecture. Furthermore, experiments conducted on artificial and real-world datasets indicate that \our{} outperforms existing approaches, especially for high noise in a candidate set.
title ProPML: Probability Partial Multi-label Learning
topic Machine Learning
url https://arxiv.org/abs/2403.07603