Non-negative residual matrix factorization: Problem definition, fast solutions, and applications

Hanghang Tong, Ching Yung Lin

Research output: Contribution to journalArticle

4 Citations (Scopus)

Abstract

Matrix factorization is a very powerful tool to find graph patterns, e.g. communities, anomalies, etc. A recent trend is to improve the usability of the discovered graph patterns, by encoding some interpretation-friendly properties (e.g., non-negativity, sparseness, etc) in the factorization. Most, if not all, of these methods are tailored for the task of community detection.We propose NrMF, a non-negative residual matrix factorization framework, aiming to improve the interpretation for graph anomaly detection. We present two optimization formations and their corresponding optimization solutions. Our method can naturally capture abnormal behaviors on graphs. We further generalize it to admit sparse constrains in the residual matrix. The effectiveness and efficiency of the proposed algorithms are analyzed, showing that our algorithm (i) leads to a local optima; and (ii) scales to large graphs. The experimental results on several data sets validate its effectiveness as well as efficiency.

Original languageEnglish (US)
Pages (from-to)3-15
Number of pages13
JournalStatistical Analysis and Data Mining
Volume5
Issue number1
DOIs
StatePublished - Feb 2012
Externally publishedYes

Fingerprint

Matrix Factorization
Factorization
Non-negative
Graph in graph theory
Community Detection
Optimization
Nonnegativity
Anomaly Detection
Usability
Anomaly
Encoding
Generalise
Experimental Results

Keywords

  • Graph mining
  • Matrix factorization
  • Non-negativity
  • Scalability

ASJC Scopus subject areas

  • Information Systems
  • Computer Science Applications
  • Analysis

Cite this

Non-negative residual matrix factorization : Problem definition, fast solutions, and applications. / Tong, Hanghang; Lin, Ching Yung.

In: Statistical Analysis and Data Mining, Vol. 5, No. 1, 02.2012, p. 3-15.

Research output: Contribution to journalArticle

@article{c4c611b4e0794490a26fc0dc46a884cc,
title = "Non-negative residual matrix factorization: Problem definition, fast solutions, and applications",
abstract = "Matrix factorization is a very powerful tool to find graph patterns, e.g. communities, anomalies, etc. A recent trend is to improve the usability of the discovered graph patterns, by encoding some interpretation-friendly properties (e.g., non-negativity, sparseness, etc) in the factorization. Most, if not all, of these methods are tailored for the task of community detection.We propose NrMF, a non-negative residual matrix factorization framework, aiming to improve the interpretation for graph anomaly detection. We present two optimization formations and their corresponding optimization solutions. Our method can naturally capture abnormal behaviors on graphs. We further generalize it to admit sparse constrains in the residual matrix. The effectiveness and efficiency of the proposed algorithms are analyzed, showing that our algorithm (i) leads to a local optima; and (ii) scales to large graphs. The experimental results on several data sets validate its effectiveness as well as efficiency.",
keywords = "Graph mining, Matrix factorization, Non-negativity, Scalability",
author = "Hanghang Tong and Lin, {Ching Yung}",
year = "2012",
month = "2",
doi = "10.1002/sam.10148",
language = "English (US)",
volume = "5",
pages = "3--15",
journal = "Statistical Analysis and Data Mining",
issn = "1932-1864",
publisher = "John Wiley and Sons Inc.",
number = "1",

}

TY - JOUR

T1 - Non-negative residual matrix factorization

T2 - Problem definition, fast solutions, and applications

AU - Tong, Hanghang

AU - Lin, Ching Yung

PY - 2012/2

Y1 - 2012/2

N2 - Matrix factorization is a very powerful tool to find graph patterns, e.g. communities, anomalies, etc. A recent trend is to improve the usability of the discovered graph patterns, by encoding some interpretation-friendly properties (e.g., non-negativity, sparseness, etc) in the factorization. Most, if not all, of these methods are tailored for the task of community detection.We propose NrMF, a non-negative residual matrix factorization framework, aiming to improve the interpretation for graph anomaly detection. We present two optimization formations and their corresponding optimization solutions. Our method can naturally capture abnormal behaviors on graphs. We further generalize it to admit sparse constrains in the residual matrix. The effectiveness and efficiency of the proposed algorithms are analyzed, showing that our algorithm (i) leads to a local optima; and (ii) scales to large graphs. The experimental results on several data sets validate its effectiveness as well as efficiency.

AB - Matrix factorization is a very powerful tool to find graph patterns, e.g. communities, anomalies, etc. A recent trend is to improve the usability of the discovered graph patterns, by encoding some interpretation-friendly properties (e.g., non-negativity, sparseness, etc) in the factorization. Most, if not all, of these methods are tailored for the task of community detection.We propose NrMF, a non-negative residual matrix factorization framework, aiming to improve the interpretation for graph anomaly detection. We present two optimization formations and their corresponding optimization solutions. Our method can naturally capture abnormal behaviors on graphs. We further generalize it to admit sparse constrains in the residual matrix. The effectiveness and efficiency of the proposed algorithms are analyzed, showing that our algorithm (i) leads to a local optima; and (ii) scales to large graphs. The experimental results on several data sets validate its effectiveness as well as efficiency.

KW - Graph mining

KW - Matrix factorization

KW - Non-negativity

KW - Scalability

UR - http://www.scopus.com/inward/record.url?scp=84863154938&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84863154938&partnerID=8YFLogxK

U2 - 10.1002/sam.10148

DO - 10.1002/sam.10148

M3 - Article

AN - SCOPUS:84863154938

VL - 5

SP - 3

EP - 15

JO - Statistical Analysis and Data Mining

JF - Statistical Analysis and Data Mining

SN - 1932-1864

IS - 1

ER -