A reconstruction error based framework for multi-label and multi-view learning

Buyue Qian, Xiang Wang, Jieping Ye, Ian Davidson

Research output: Contribution to journalArticle

18 Citations (Scopus)

Abstract

A significant challenge to make learning techniques more suitable for general purpose use is to move beyond i) complete supervision, ii) low dimensional data, iii) a single label and single view per instance. Solving these challenges allows working with complex learning problems that are typically high dimensional with multiple (but possibly incomplete) labelings and views. While other work has addressed each of these problems separately, in this paper we show how to address them together, namely semi-supervised dimension reduction for multi-label and multi-view learning (SSDR-MML), which performs optimization for dimension reduction and label inference in semi-supervised setting. The proposed framework is designed to handle both multi-label and multi-view learningsettings, and can be easily extended to many useful applications. Our formulation has a number of advantages. We explicitly model the information combining mechanism as a data structure (a weight/nearest-neighbor matrix) which allows investigating fundamentalquestions in multi-label and multi-view learning. We address one such question by presenting a general measure to quantify thesuccess of simultaneous learning of multiple labels or views. We empirically demonstrate the usefulness of our SSDR-MML approach, and show that it can outperform many state-of-the-art baseline methods.

Original languageEnglish (US)
Article number2339860
Pages (from-to)594-607
Number of pages14
JournalIEEE Transactions on Knowledge and Data Engineering
Volume27
Issue number3
DOIs
StatePublished - Mar 1 2015

Fingerprint

Labels
Labeling
Data structures

Keywords

  • dimension reduction
  • multi-label learning
  • multi-view learning
  • reconstruction error
  • Semi-supervised learning

ASJC Scopus subject areas

  • Computational Theory and Mathematics
  • Information Systems
  • Computer Science Applications

Cite this

A reconstruction error based framework for multi-label and multi-view learning. / Qian, Buyue; Wang, Xiang; Ye, Jieping; Davidson, Ian.

In: IEEE Transactions on Knowledge and Data Engineering, Vol. 27, No. 3, 2339860, 01.03.2015, p. 594-607.

Research output: Contribution to journalArticle

Qian, Buyue ; Wang, Xiang ; Ye, Jieping ; Davidson, Ian. / A reconstruction error based framework for multi-label and multi-view learning. In: IEEE Transactions on Knowledge and Data Engineering. 2015 ; Vol. 27, No. 3. pp. 594-607.
@article{ca5a494bf1d04e1c8d3de0e071e8c99e,
title = "A reconstruction error based framework for multi-label and multi-view learning",
abstract = "A significant challenge to make learning techniques more suitable for general purpose use is to move beyond i) complete supervision, ii) low dimensional data, iii) a single label and single view per instance. Solving these challenges allows working with complex learning problems that are typically high dimensional with multiple (but possibly incomplete) labelings and views. While other work has addressed each of these problems separately, in this paper we show how to address them together, namely semi-supervised dimension reduction for multi-label and multi-view learning (SSDR-MML), which performs optimization for dimension reduction and label inference in semi-supervised setting. The proposed framework is designed to handle both multi-label and multi-view learningsettings, and can be easily extended to many useful applications. Our formulation has a number of advantages. We explicitly model the information combining mechanism as a data structure (a weight/nearest-neighbor matrix) which allows investigating fundamentalquestions in multi-label and multi-view learning. We address one such question by presenting a general measure to quantify thesuccess of simultaneous learning of multiple labels or views. We empirically demonstrate the usefulness of our SSDR-MML approach, and show that it can outperform many state-of-the-art baseline methods.",
keywords = "dimension reduction, multi-label learning, multi-view learning, reconstruction error, Semi-supervised learning",
author = "Buyue Qian and Xiang Wang and Jieping Ye and Ian Davidson",
year = "2015",
month = "3",
day = "1",
doi = "10.1109/TKDE.2014.2339860",
language = "English (US)",
volume = "27",
pages = "594--607",
journal = "IEEE Transactions on Knowledge and Data Engineering",
issn = "1041-4347",
publisher = "IEEE Computer Society",
number = "3",

}

TY - JOUR

T1 - A reconstruction error based framework for multi-label and multi-view learning

AU - Qian, Buyue

AU - Wang, Xiang

AU - Ye, Jieping

AU - Davidson, Ian

PY - 2015/3/1

Y1 - 2015/3/1

N2 - A significant challenge to make learning techniques more suitable for general purpose use is to move beyond i) complete supervision, ii) low dimensional data, iii) a single label and single view per instance. Solving these challenges allows working with complex learning problems that are typically high dimensional with multiple (but possibly incomplete) labelings and views. While other work has addressed each of these problems separately, in this paper we show how to address them together, namely semi-supervised dimension reduction for multi-label and multi-view learning (SSDR-MML), which performs optimization for dimension reduction and label inference in semi-supervised setting. The proposed framework is designed to handle both multi-label and multi-view learningsettings, and can be easily extended to many useful applications. Our formulation has a number of advantages. We explicitly model the information combining mechanism as a data structure (a weight/nearest-neighbor matrix) which allows investigating fundamentalquestions in multi-label and multi-view learning. We address one such question by presenting a general measure to quantify thesuccess of simultaneous learning of multiple labels or views. We empirically demonstrate the usefulness of our SSDR-MML approach, and show that it can outperform many state-of-the-art baseline methods.

AB - A significant challenge to make learning techniques more suitable for general purpose use is to move beyond i) complete supervision, ii) low dimensional data, iii) a single label and single view per instance. Solving these challenges allows working with complex learning problems that are typically high dimensional with multiple (but possibly incomplete) labelings and views. While other work has addressed each of these problems separately, in this paper we show how to address them together, namely semi-supervised dimension reduction for multi-label and multi-view learning (SSDR-MML), which performs optimization for dimension reduction and label inference in semi-supervised setting. The proposed framework is designed to handle both multi-label and multi-view learningsettings, and can be easily extended to many useful applications. Our formulation has a number of advantages. We explicitly model the information combining mechanism as a data structure (a weight/nearest-neighbor matrix) which allows investigating fundamentalquestions in multi-label and multi-view learning. We address one such question by presenting a general measure to quantify thesuccess of simultaneous learning of multiple labels or views. We empirically demonstrate the usefulness of our SSDR-MML approach, and show that it can outperform many state-of-the-art baseline methods.

KW - dimension reduction

KW - multi-label learning

KW - multi-view learning

KW - reconstruction error

KW - Semi-supervised learning

UR - http://www.scopus.com/inward/record.url?scp=84922840934&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84922840934&partnerID=8YFLogxK

U2 - 10.1109/TKDE.2014.2339860

DO - 10.1109/TKDE.2014.2339860

M3 - Article

VL - 27

SP - 594

EP - 607

JO - IEEE Transactions on Knowledge and Data Engineering

JF - IEEE Transactions on Knowledge and Data Engineering

SN - 1041-4347

IS - 3

M1 - 2339860

ER -