Learning and Storing the Parts of Objects: IMF

Ruairí de Fréin

doi:10.1109/MLSP.2014.6958926

Learning and Storing the Parts of Objects: IMF

Ruairí de Fréin

Department of Computing and Mathematics

Research output: Contribution to conference › Paper › peer-review

6 Citations (Scopus)

Abstract

A central concern for many learning algorithms is how to efficiently store what the algorithm has learned. An algorithm for the compression of Nonnegative Matrix Factorizations is presented. Compression is achieved by embedding the factorization in an encoding routine. Its performance is investigated using two standard test images, Peppers and Barbara. The compression ratio (18:1) achieved by the proposed Matrix Factorization improves the storage-ability of Nonnegative Matrix Factorizations without significantly degrading accuracy (≈ 1-3dB degradation is introduced). We learn as before, but storage is cheaper.

Original language	English
DOIs	https://doi.org/10.1109/MLSP.2014.6958926
Publication status	Published - 2014
Event	IEEE International Workshop on Machine Learning for Signal Processing - Reims, France, Reims, France Duration: 01 Jan 2014 → … http://mlsp2014.conwiz.dk/home.htm#.VDE3necjEbI

Conference

Conference	IEEE International Workshop on Machine Learning for Signal Processing
City	Reims, France
Period	01/01/2014 → …
Internet address	http://mlsp2014.conwiz.dk/home.htm#.VDE3necjEbI

Keywords

Adaptive algorithms.
Data-driven adaptive systems and models;
Learning theory and techniques;
compression;
matrix factorization;

Access to Document

10.1109/MLSP.2014.6958926

Learning and Storing the Parts of Objects: IMFOther version, 0 bytes

http://repository.wit.ie/id/document/4624

Cite this

@conference{cd2b1ce9db4e4b15b26ce986b50473e5,

title = "Learning and Storing the Parts of Objects: IMF",

abstract = "A central concern for many learning algorithms is how to efficiently store what the algorithm has learned. An algorithm for the compression of Nonnegative Matrix Factorizations is presented. Compression is achieved by embedding the factorization in an encoding routine. Its performance is investigated using two standard test images, Peppers and Barbara. The compression ratio (18:1) achieved by the proposed Matrix Factorization improves the storage-ability of Nonnegative Matrix Factorizations without significantly degrading accuracy (≈ 1-3dB degradation is introduced). We learn as before, but storage is cheaper.",

keywords = "Adaptive algorithms., Data-driven adaptive systems and models;, Learning theory and techniques;, compression;, matrix factorization;",

author = "{de Fr{\'e}in}, Ruair{\'i}",

year = "2014",

doi = "10.1109/MLSP.2014.6958926",

language = "English",

note = "IEEE International Workshop on Machine Learning for Signal Processing ; Conference date: 01-01-2014",

url = "http://mlsp2014.conwiz.dk/home.htm#.VDE3necjEbI",

}

TY - CONF

T1 - Learning and Storing the Parts of Objects: IMF

AU - de Fréin, Ruairí

PY - 2014

Y1 - 2014

N2 - A central concern for many learning algorithms is how to efficiently store what the algorithm has learned. An algorithm for the compression of Nonnegative Matrix Factorizations is presented. Compression is achieved by embedding the factorization in an encoding routine. Its performance is investigated using two standard test images, Peppers and Barbara. The compression ratio (18:1) achieved by the proposed Matrix Factorization improves the storage-ability of Nonnegative Matrix Factorizations without significantly degrading accuracy (≈ 1-3dB degradation is introduced). We learn as before, but storage is cheaper.

AB - A central concern for many learning algorithms is how to efficiently store what the algorithm has learned. An algorithm for the compression of Nonnegative Matrix Factorizations is presented. Compression is achieved by embedding the factorization in an encoding routine. Its performance is investigated using two standard test images, Peppers and Barbara. The compression ratio (18:1) achieved by the proposed Matrix Factorization improves the storage-ability of Nonnegative Matrix Factorizations without significantly degrading accuracy (≈ 1-3dB degradation is introduced). We learn as before, but storage is cheaper.

KW - Adaptive algorithms.

KW - Data-driven adaptive systems and models;

KW - Learning theory and techniques;

KW - compression;

KW - matrix factorization;

U2 - 10.1109/MLSP.2014.6958926

DO - 10.1109/MLSP.2014.6958926

M3 - Paper

T2 - IEEE International Workshop on Machine Learning for Signal Processing

Y2 - 1 January 2014

ER -

Learning and Storing the Parts of Objects: IMF

Abstract

Conference

Keywords

Access to Document

Fingerprint

Cite this