Modelling dense relational data

Tue Herlau; Morten Mørup; Mikkel Nørgaard Schmidt; Lars Kai Hansen

doi:10.1109/MLSP.2012.6349747

Modelling dense relational data

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

Abstract

Relational modelling classically consider sparse and discrete data. Measures of influence computed pairwise between temporal sources naturally give rise to dense continuous-valued matrices, for instance p-values from Granger causality. Due to asymmetry or lack of positive definiteness they are not naturally suited for kernel K-means. We propose a generative Bayesian model for dense matrices which generalize kernel K-means to consider off-diagonal interactions in matrices of interactions, and demonstrate its ability to detect structure on both artificial data and two real data sets.

Original language	English
Title of host publication	2012 IEEE International Workshop on Machine Learning for Signal Processing (MLSP)
Number of pages	6
Place of Publication	978-1-4673-1025-3
Publisher	IEEE
Publication date	2012
ISBN (Print)	978-1-4673-1024-6
DOIs	https://doi.org/10.1109/MLSP.2012.6349747
Publication status	Published - 2012
Event	2012 IEEE International Workshop on Machine Learning for Signal Processing - Santander, Spain Duration: 23 Oct 2012 → 26 Oct 2012 Conference number: 22 https://ieeexplore.ieee.org/xpl/conhome/6335571/proceeding

Conference

Conference	2012 IEEE International Workshop on Machine Learning for Signal Processing
Number	22
Country/Territory	Spain
City	Santander
Period	23/10/2012 → 26/10/2012
Internet address	https://ieeexplore.ieee.org/xpl/conhome/6335571/proceeding

Series	Machine Learning for Signal Processing
ISSN	1551-2541

Keywords

Relational Modelling
Non-parametrics
Infinite Relational Model
Granger Causality

Access to Document

10.1109/MLSP.2012.6349747

OpenUrl availability

Full text

Cite this

@inproceedings{cefb9c602b4a48b3b62e8801f4a2716b,

title = "Modelling dense relational data",

abstract = "Relational modelling classically consider sparse and discrete data. Measures of influence computed pairwise between temporal sources naturally give rise to dense continuous-valued matrices, for instance p-values from Granger causality. Due to asymmetry or lack of positive definiteness they are not naturally suited for kernel K-means. We propose a generative Bayesian model for dense matrices which generalize kernel K-means to consider off-diagonal interactions in matrices of interactions, and demonstrate its ability to detect structure on both artificial data and two real data sets.",

keywords = "Relational Modelling, Non-parametrics, Infinite Relational Model, Granger Causality",

author = "Tue Herlau and Morten M{\o}rup and Schmidt, \{Mikkel N{\o}rgaard\} and Hansen, \{Lars Kai\}",

year = "2012",

doi = "10.1109/MLSP.2012.6349747",

language = "English",

isbn = "978-1-4673-1024-6",

series = "Machine Learning for Signal Processing",

booktitle = "2012 IEEE International Workshop on Machine Learning for Signal Processing (MLSP)",

publisher = "IEEE",

address = "United States",

note = "2012 IEEE International Workshop on Machine Learning for Signal Processing, MLSP 2012 ; Conference date: 23-10-2012 Through 26-10-2012",

url = "https://ieeexplore.ieee.org/xpl/conhome/6335571/proceeding",

}

Herlau, T , Mørup, M , Schmidt, MN & Hansen, LK 2012, Modelling dense relational data. in 2012 IEEE International Workshop on Machine Learning for Signal Processing (MLSP). IEEE, 978-1-4673-1025-3, Machine Learning for Signal Processing, 2012 IEEE International Workshop on Machine Learning for Signal Processing, Santander, Spain, 23/10/2012. https://doi.org/10.1109/MLSP.2012.6349747

Modelling dense relational data. / Herlau, Tue ; Mørup, Morten ; Schmidt, Mikkel Nørgaard et al.
2012 IEEE International Workshop on Machine Learning for Signal Processing (MLSP). 978-1-4673-1025-3: IEEE, 2012. (Machine Learning for Signal Processing).

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

TY - GEN

T1 - Modelling dense relational data

AU - Herlau, Tue

AU - Mørup, Morten

AU - Schmidt, Mikkel Nørgaard

AU - Hansen, Lars Kai

N1 - Conference code: 22

PY - 2012

Y1 - 2012

N2 - Relational modelling classically consider sparse and discrete data. Measures of influence computed pairwise between temporal sources naturally give rise to dense continuous-valued matrices, for instance p-values from Granger causality. Due to asymmetry or lack of positive definiteness they are not naturally suited for kernel K-means. We propose a generative Bayesian model for dense matrices which generalize kernel K-means to consider off-diagonal interactions in matrices of interactions, and demonstrate its ability to detect structure on both artificial data and two real data sets.

AB - Relational modelling classically consider sparse and discrete data. Measures of influence computed pairwise between temporal sources naturally give rise to dense continuous-valued matrices, for instance p-values from Granger causality. Due to asymmetry or lack of positive definiteness they are not naturally suited for kernel K-means. We propose a generative Bayesian model for dense matrices which generalize kernel K-means to consider off-diagonal interactions in matrices of interactions, and demonstrate its ability to detect structure on both artificial data and two real data sets.

KW - Relational Modelling

KW - Non-parametrics

KW - Infinite Relational Model

KW - Granger Causality

U2 - 10.1109/MLSP.2012.6349747

DO - 10.1109/MLSP.2012.6349747

M3 - Article in proceedings

SN - 978-1-4673-1024-6

T3 - Machine Learning for Signal Processing

BT - 2012 IEEE International Workshop on Machine Learning for Signal Processing (MLSP)

PB - IEEE

CY - 978-1-4673-1025-3

T2 - 2012 IEEE International Workshop on Machine Learning for Signal Processing

Y2 - 23 October 2012 through 26 October 2012

ER -

Modelling dense relational data

Abstract

Conference

Keywords

Access to Document

OpenUrl availability

Fingerprint

Cite this