On flat versus hierarchical classification in large-scale taxonomies

Rohit Babbar; Ioannis Partalas; Eric Gaussier; Massih Reza Amini

On flat versus hierarchical classification in large-scale taxonomies

Rohit Babbar, Ioannis Partalas, Eric Gaussier, Massih Reza Amini

Department of Computer Science

Institut NEEL

Research output: Contribution to journal › Conference article › peer-review

58 Citations (SciVal)

Abstract

We study in this paper flat and hierarchical classification strategies in the context of large-scale taxonomies. To this end, we first propose a multiclass, hierarchical data dependent bound on the generalization error of classifiers deployed in large-scale taxonomies. This bound provides an explanation to several empirical results reported in the literature, related to the performance of flat and hierarchical classifiers. We then introduce another type of bound targeting the approximation error of a family of classifiers, and derive from it features used in a meta-classifier to decide which nodes to prune (or flatten) in a large-scale taxonomy. We finally illustrate the theoretical developments through several experiments conducted on two widely used taxonomies.

Original language	English
Pages (from-to)	1-9
Number of pages	9
Journal	Advances in Neural Information Processing Systems
Publication status	Published - 5 Dec 2013
Event	27th Annual Conference on Neural Information Processing Systems, NIPS 2013 - Lake Tahoe, NV, USA United States Duration: 5 Dec 2013 → 10 Dec 2013

Bibliographical note

Funding information:
This work was supported in part by the ANR project Class-Y, the Mastodons project Garguantua, the LabEx PERSYVAL-Lab ANR-11-LABX-0025 and the European project BioASQ (grant agreement no. 318652).

ASJC Scopus subject areas

Computer Networks and Communications
Information Systems
Signal Processing

Access to Document

https://papers.nips.cc/paper_files/paper/2013/hash/cbb6a3b884f4f88b3a8e3d44c636cbd8-Abstract.htmlLicence: Unspecified

Cite this

@article{458982524dc046edbdb97440ac56f176,

title = "On flat versus hierarchical classification in large-scale taxonomies",

abstract = "We study in this paper flat and hierarchical classification strategies in the context of large-scale taxonomies. To this end, we first propose a multiclass, hierarchical data dependent bound on the generalization error of classifiers deployed in large-scale taxonomies. This bound provides an explanation to several empirical results reported in the literature, related to the performance of flat and hierarchical classifiers. We then introduce another type of bound targeting the approximation error of a family of classifiers, and derive from it features used in a meta-classifier to decide which nodes to prune (or flatten) in a large-scale taxonomy. We finally illustrate the theoretical developments through several experiments conducted on two widely used taxonomies.",

author = "Rohit Babbar and Ioannis Partalas and Eric Gaussier and Amini, {Massih Reza}",

note = "Funding information: This work was supported in part by the ANR project Class-Y, the Mastodons project Garguantua, the LabEx PERSYVAL-Lab ANR-11-LABX-0025 and the European project BioASQ (grant agreement no. 318652).; 27th Annual Conference on Neural Information Processing Systems, NIPS 2013 ; Conference date: 05-12-2013 Through 10-12-2013",

year = "2013",

month = dec,

day = "5",

language = "English",

pages = "1--9",

journal = "Advances in Neural Information Processing Systems",

issn = "1049-5258",

publisher = "Neural information processing systems foundation",

}

TY - JOUR

T1 - On flat versus hierarchical classification in large-scale taxonomies

AU - Babbar, Rohit

AU - Partalas, Ioannis

AU - Gaussier, Eric

AU - Amini, Massih Reza

N1 - Funding information: This work was supported in part by the ANR project Class-Y, the Mastodons project Garguantua, the LabEx PERSYVAL-Lab ANR-11-LABX-0025 and the European project BioASQ (grant agreement no. 318652).

PY - 2013/12/5

Y1 - 2013/12/5

N2 - We study in this paper flat and hierarchical classification strategies in the context of large-scale taxonomies. To this end, we first propose a multiclass, hierarchical data dependent bound on the generalization error of classifiers deployed in large-scale taxonomies. This bound provides an explanation to several empirical results reported in the literature, related to the performance of flat and hierarchical classifiers. We then introduce another type of bound targeting the approximation error of a family of classifiers, and derive from it features used in a meta-classifier to decide which nodes to prune (or flatten) in a large-scale taxonomy. We finally illustrate the theoretical developments through several experiments conducted on two widely used taxonomies.

AB - We study in this paper flat and hierarchical classification strategies in the context of large-scale taxonomies. To this end, we first propose a multiclass, hierarchical data dependent bound on the generalization error of classifiers deployed in large-scale taxonomies. This bound provides an explanation to several empirical results reported in the literature, related to the performance of flat and hierarchical classifiers. We then introduce another type of bound targeting the approximation error of a family of classifiers, and derive from it features used in a meta-classifier to decide which nodes to prune (or flatten) in a large-scale taxonomy. We finally illustrate the theoretical developments through several experiments conducted on two widely used taxonomies.

UR - http://www.scopus.com/inward/record.url?scp=84899031281&partnerID=8YFLogxK

M3 - Conference article

AN - SCOPUS:84899031281

SN - 1049-5258

SP - 1

EP - 9

JO - Advances in Neural Information Processing Systems

JF - Advances in Neural Information Processing Systems

T2 - 27th Annual Conference on Neural Information Processing Systems, NIPS 2013

Y2 - 5 December 2013 through 10 December 2013

ER -

On flat versus hierarchical classification in large-scale taxonomies

Abstract

Bibliographical note

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this