Compressive mining: fast and optimal data mining in the compressed domain

Michail Vlachos, Nikolaos Freris, Anastasios Kyrillidis

Research output: Contribution to journalArticle

Abstract

Real-world data typically contain repeated and periodic patterns. This suggests that they can be effectively represented and compressed using only a few coefficients of an appropriate basis (e.g., Fourier and wavelets). However, distance estimation when the data are represented using different sets of coefficients is still a largely unexplored area. This work studies the optimization problems related to obtaining the tightest lower/upper bound on Euclidean distances when each data object is potentially compressed using a different set of orthonormal coefficients. Our technique leads to tighter distance estimates, which translates into more accurate search, learning and mining operations directly in the compressed domain. We formulate the problem of estimating lower/upper distance bounds as an optimization problem. We establish the properties of optimal solutions and leverage the theoretical analysis to develop a fast algorithm to obtain an exact solution to the problem. The suggested solution provides the tightest estimation of the $$L_2$$L2-norm or the correlation. We show that typical data analysis operations, such as $$k$$k-nearest-neighbor search or k-Means clustering, can operate more accurately using the proposed compression and distance reconstruction technique. We compare it with many other prevalent compression and reconstruction techniques, including random projections and PCA-based techniques. We highlight a surprising result, namely that when the data are highly sparse in some basis, our technique may even outperform PCA-based compression. The contributions of this work are generic as our methodology is applicable to any sequential or high-dimensional data as well as to any orthogonal data transformation used for the underlying data compression scheme.

Original languageEnglish (US)
JournalVLDB Journal
Volume24
Issue number1
DOIs
StatePublished - Jan 17 2015

Fingerprint

Data mining
Data compression
Nearest neighbor search

Keywords

  • Compressive sensing
  • Convex optimization
  • Data compression
  • Fourier
  • Waterfilling algorithm
  • wavelets

ASJC Scopus subject areas

  • Information Systems
  • Hardware and Architecture

Cite this

Compressive mining : fast and optimal data mining in the compressed domain. / Vlachos, Michail; Freris, Nikolaos; Kyrillidis, Anastasios.

In: VLDB Journal, Vol. 24, No. 1, 17.01.2015.

Research output: Contribution to journalArticle

Vlachos, Michail ; Freris, Nikolaos ; Kyrillidis, Anastasios. / Compressive mining : fast and optimal data mining in the compressed domain. In: VLDB Journal. 2015 ; Vol. 24, No. 1.
@article{d971b40ed1e443e3af7f5fbaf90636af,
title = "Compressive mining: fast and optimal data mining in the compressed domain",
abstract = "Real-world data typically contain repeated and periodic patterns. This suggests that they can be effectively represented and compressed using only a few coefficients of an appropriate basis (e.g., Fourier and wavelets). However, distance estimation when the data are represented using different sets of coefficients is still a largely unexplored area. This work studies the optimization problems related to obtaining the tightest lower/upper bound on Euclidean distances when each data object is potentially compressed using a different set of orthonormal coefficients. Our technique leads to tighter distance estimates, which translates into more accurate search, learning and mining operations directly in the compressed domain. We formulate the problem of estimating lower/upper distance bounds as an optimization problem. We establish the properties of optimal solutions and leverage the theoretical analysis to develop a fast algorithm to obtain an exact solution to the problem. The suggested solution provides the tightest estimation of the $$L_2$$L2-norm or the correlation. We show that typical data analysis operations, such as $$k$$k-nearest-neighbor search or k-Means clustering, can operate more accurately using the proposed compression and distance reconstruction technique. We compare it with many other prevalent compression and reconstruction techniques, including random projections and PCA-based techniques. We highlight a surprising result, namely that when the data are highly sparse in some basis, our technique may even outperform PCA-based compression. The contributions of this work are generic as our methodology is applicable to any sequential or high-dimensional data as well as to any orthogonal data transformation used for the underlying data compression scheme.",
keywords = "Compressive sensing, Convex optimization, Data compression, Fourier, Waterfilling algorithm, wavelets",
author = "Michail Vlachos and Nikolaos Freris and Anastasios Kyrillidis",
year = "2015",
month = "1",
day = "17",
doi = "10.1007/s00778-014-0360-3",
language = "English (US)",
volume = "24",
journal = "VLDB Journal",
issn = "1066-8888",
publisher = "Springer New York",
number = "1",

}

TY - JOUR

T1 - Compressive mining

T2 - fast and optimal data mining in the compressed domain

AU - Vlachos, Michail

AU - Freris, Nikolaos

AU - Kyrillidis, Anastasios

PY - 2015/1/17

Y1 - 2015/1/17

N2 - Real-world data typically contain repeated and periodic patterns. This suggests that they can be effectively represented and compressed using only a few coefficients of an appropriate basis (e.g., Fourier and wavelets). However, distance estimation when the data are represented using different sets of coefficients is still a largely unexplored area. This work studies the optimization problems related to obtaining the tightest lower/upper bound on Euclidean distances when each data object is potentially compressed using a different set of orthonormal coefficients. Our technique leads to tighter distance estimates, which translates into more accurate search, learning and mining operations directly in the compressed domain. We formulate the problem of estimating lower/upper distance bounds as an optimization problem. We establish the properties of optimal solutions and leverage the theoretical analysis to develop a fast algorithm to obtain an exact solution to the problem. The suggested solution provides the tightest estimation of the $$L_2$$L2-norm or the correlation. We show that typical data analysis operations, such as $$k$$k-nearest-neighbor search or k-Means clustering, can operate more accurately using the proposed compression and distance reconstruction technique. We compare it with many other prevalent compression and reconstruction techniques, including random projections and PCA-based techniques. We highlight a surprising result, namely that when the data are highly sparse in some basis, our technique may even outperform PCA-based compression. The contributions of this work are generic as our methodology is applicable to any sequential or high-dimensional data as well as to any orthogonal data transformation used for the underlying data compression scheme.

AB - Real-world data typically contain repeated and periodic patterns. This suggests that they can be effectively represented and compressed using only a few coefficients of an appropriate basis (e.g., Fourier and wavelets). However, distance estimation when the data are represented using different sets of coefficients is still a largely unexplored area. This work studies the optimization problems related to obtaining the tightest lower/upper bound on Euclidean distances when each data object is potentially compressed using a different set of orthonormal coefficients. Our technique leads to tighter distance estimates, which translates into more accurate search, learning and mining operations directly in the compressed domain. We formulate the problem of estimating lower/upper distance bounds as an optimization problem. We establish the properties of optimal solutions and leverage the theoretical analysis to develop a fast algorithm to obtain an exact solution to the problem. The suggested solution provides the tightest estimation of the $$L_2$$L2-norm or the correlation. We show that typical data analysis operations, such as $$k$$k-nearest-neighbor search or k-Means clustering, can operate more accurately using the proposed compression and distance reconstruction technique. We compare it with many other prevalent compression and reconstruction techniques, including random projections and PCA-based techniques. We highlight a surprising result, namely that when the data are highly sparse in some basis, our technique may even outperform PCA-based compression. The contributions of this work are generic as our methodology is applicable to any sequential or high-dimensional data as well as to any orthogonal data transformation used for the underlying data compression scheme.

KW - Compressive sensing

KW - Convex optimization

KW - Data compression

KW - Fourier

KW - Waterfilling algorithm

KW - wavelets

UR - http://www.scopus.com/inward/record.url?scp=84938290168&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84938290168&partnerID=8YFLogxK

U2 - 10.1007/s00778-014-0360-3

DO - 10.1007/s00778-014-0360-3

M3 - Article

AN - SCOPUS:84938290168

VL - 24

JO - VLDB Journal

JF - VLDB Journal

SN - 1066-8888

IS - 1

ER -