Multiresolution matrix factorization

Risi Kondor; Nedelina Teneva; Vikas K. Garg

Multiresolution matrix factorization

Risi Kondor, Nedelina Teneva, Vikas K. Garg

Not published at Aalto University

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

18 Citations (Scopus)

Abstract

Large matrices appearing in machine learning problems often have complex hierarchical structures that go beyond what can be found by traditional linear algebra tools, such as eigende-composition. Inspired by ideas from multiresolution analysis, this paper introduces a new notion of matrix factorization that can capture structure in matrices at multiple different scales. The resulting Multiresolution Matrix Factorizations (MMFs) not only provide a wavelet basis for sparse approximation, but can also be used for matrix compression (similar to Nyström approximations) and as a prior for matrix completion.

Original language	English
Title of host publication	31st International Conference on Machine Learning, ICML 2014
Publisher	International Machine Learning Society
Pages	3591-3601
Number of pages	11
ISBN (Electronic)	9781634393973
Publication status	Published - 2014
MoE publication type	A4 Conference publication
Event	International Conference on Machine Learning - Beijing, China Duration: 21 Jun 2014 → 26 Jun 2014 Conference number: 31

Conference

Conference	International Conference on Machine Learning
Abbreviated title	ICML
Country/Territory	China
City	Beijing
Period	21/06/2014 → 26/06/2014

Cite this

@inproceedings{f88a0d05205a4cada4ef9fea8c2a3d57,

title = "Multiresolution matrix factorization",

abstract = "Large matrices appearing in machine learning problems often have complex hierarchical structures that go beyond what can be found by traditional linear algebra tools, such as eigende-composition. Inspired by ideas from multiresolution analysis, this paper introduces a new notion of matrix factorization that can capture structure in matrices at multiple different scales. The resulting Multiresolution Matrix Factorizations (MMFs) not only provide a wavelet basis for sparse approximation, but can also be used for matrix compression (similar to Nystr{\"o}m approximations) and as a prior for matrix completion.",

author = "Risi Kondor and Nedelina Teneva and Garg, {Vikas K.}",

year = "2014",

language = "English",

pages = "3591--3601",

booktitle = "31st International Conference on Machine Learning, ICML 2014",

publisher = "International Machine Learning Society ",

address = "United States",

note = "International Conference on Machine Learning, ICML ; Conference date: 21-06-2014 Through 26-06-2014",

}

TY - GEN

T1 - Multiresolution matrix factorization

AU - Kondor, Risi

AU - Teneva, Nedelina

AU - Garg, Vikas K.

N1 - Conference code: 31

PY - 2014

Y1 - 2014

N2 - Large matrices appearing in machine learning problems often have complex hierarchical structures that go beyond what can be found by traditional linear algebra tools, such as eigende-composition. Inspired by ideas from multiresolution analysis, this paper introduces a new notion of matrix factorization that can capture structure in matrices at multiple different scales. The resulting Multiresolution Matrix Factorizations (MMFs) not only provide a wavelet basis for sparse approximation, but can also be used for matrix compression (similar to Nyström approximations) and as a prior for matrix completion.

AB - Large matrices appearing in machine learning problems often have complex hierarchical structures that go beyond what can be found by traditional linear algebra tools, such as eigende-composition. Inspired by ideas from multiresolution analysis, this paper introduces a new notion of matrix factorization that can capture structure in matrices at multiple different scales. The resulting Multiresolution Matrix Factorizations (MMFs) not only provide a wavelet basis for sparse approximation, but can also be used for matrix compression (similar to Nyström approximations) and as a prior for matrix completion.

UR - http://www.scopus.com/inward/record.url?scp=84919818836&partnerID=8YFLogxK

M3 - Conference article in proceedings

AN - SCOPUS:84919818836

SP - 3591

EP - 3601

BT - 31st International Conference on Machine Learning, ICML 2014

PB - International Machine Learning Society

T2 - International Conference on Machine Learning

Y2 - 21 June 2014 through 26 June 2014

ER -

Multiresolution matrix factorization

Abstract

Conference

Other files and links

Fingerprint

Cite this