Effects of the hierarchy in hierarchical, multi-label classification

Daisey, Katie; Brown, Steven D.

doi:10.1016/j.chemolab.2020.104177

Cited by 7 publications

(2 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…audio and images. There has also been some recent work in hierarchical multilabel text classification [70], [61], [71], [63], [72]. Local classification approaches [73] train a set of classifiers at each level of the hierarchy.…”

Section: Related Workmentioning

confidence: 99%

All Mistakes Are Not Equal: Comprehensive Hierarchy Aware Multi-label Predictions (CHAMP)

Vaswani¹,

Aggarwal²,

Netrapalli³

et al. 2022

Preprint

View full text Add to dashboard Cite

This paper considers the problem of Hierarchical Multi-Label Classification (HMC), where (i) several labels can be present for each example, and (ii) labels are related via a domain-specific hierarchy tree. Guided by the intuition that all mistakes are not equal, we present Comprehensive Hierarchy Aware Multi-label Predictions (CHAMP), a framework that penalizes a misprediction depending on its severity as per the hierarchy tree. While there have been works that apply such an idea to singlelabel classification, to the best of our knowledge, there are limited such works for multilabel classification focusing on the severity of mistakes. The key reason is that there is no clear way of quantifying the severity of a misprediction a priori in the multilabel setting. In this work, we propose a simple but effective metric to quantify the severity of a mistake in HMC, naturally leading to CHAMP. Extensive experiments on six public HMC datasets across modalities (image, audio, and text) demonstrate that incorporating hierarchical information leads to substantial gains as CHAMP improves both AUPRC (2.6% median percentage improvement) and hierarchical metrics (2.85% median percentage improvement), over stand-alone hierarchical or multilabel classification methods. Compared to standard multilabel baselines, CHAMP provides improved AUPRC in both robustness (8.87% mean percentage improvement) and less data regimes. Further, our method provides a framework to enhance existing multilabel classification algorithms with better mistakes (18.1% mean percentage improvement).Preprint. Under review.

show abstract

Section: Related Workmentioning

confidence: 99%

All Mistakes Are Not Equal: Comprehensive Hierarchy Aware Multi-label Predictions (CHAMP)

Vaswani¹,

Aggarwal²,

Netrapalli³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…. [10], Tao X., Tong S., Zhao X., Xu T., Jin B., Liu C. [11], Wehrmann J., Cerri R., Barros R. [12], Lu Junyu, Zhang Hao, Shen Zhexui, Shi Kaiyuan, Yang Liang, Xu Bo, Shaowu Zhang, Lin Hongfei [13], Daisy Cathy, Brown Stephen [14], Alaidi Noor, Riddy Chandan, Fotouhi Farshad [15] and others.…”

mentioning

confidence: 99%

Hierarchical Multi-Label Classification

Nedashkivskyi

2024

Věda a perspektivy

View full text Add to dashboard Cite

The multi-label classification (MLC) task involves assigning multiple labels to each data point based on its characteristics or features. Addressing the challenges of high or ultra-high dimensionality in multi-label data, multi-label feature selection emerges as a viable solution. Nonetheless, prevalent algorithms in this domain typically accommodate only a single data type, treat all labels uniformly, and rely on heuristic search methods. Consequently, these approaches often yield suboptimal classification accuracy, leading to inefficiencies. The most common approaches dealing with multi-label classification problems include problem transformation (PT), which aims to adapt multi-label data, making it possible to use traditional binary or multi-class classification algorithms, and algorithm adaptation (AA), which focuses on modifying algorithms used for binary or multi-class classification, allowing them to make multi-label predictions. In PT methods, algorithms solve the multi-label learning problem by transforming it into other wellestablished learning scenarios. Representative algorithms include binary relevance and classifier chains, which transform a multi-class classification (MLC) task into a binary classification task, and calibrated label ranking, which transforms an MLC task into a label ranking task. AA methods solve MLC problems by adapting popular learning methods or creating entirely new ones to deal directly with multi-label data. Within the framework of this work, a review of the methods of introducing the concept of hierarchical classification with several labels into the process of classifying textual documentation was carried out.

show abstract

A unified view of feature selection based on Hilbert-Schmidt independence criterion

Wang

Liu

2023

Chemometrics and Intelligent Laboratory Systems

View full text Add to dashboard Cite

Effects of the hierarchy in hierarchical, multi-label classification

Cited by 7 publications

References 22 publications

All Mistakes Are Not Equal: Comprehensive Hierarchy Aware Multi-label Predictions (CHAMP)

All Mistakes Are Not Equal: Comprehensive Hierarchy Aware Multi-label Predictions (CHAMP)

Hierarchical Multi-Label Classification

A unified view of feature selection based on Hilbert-Schmidt independence criterion

Contact Info

Product

Resources

About