PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

Graphical multi-way models
Ilkka Huopaniemi, Tommi Suvitaival, Matej Oresic and Samuel Kaski
In: ECML/PKDD 2010, 20 Sep - 24 Sep 2010, Barcelona.

Abstract

Multivariate multi-way ANOVA-type models are the default tools for analyzing experimental data with multiple independent covariates. However, formulating standard multi-way models is not possible when the data comes from different sources or in cases where some covariates have (partly) unknown structure, such as time with unknown alignment. The “small n, large p”, large dimensionality p with small number of samples n, settings bring further problems to the standard multivariate methods. We extend our recent graphical multi-way model to three general setups, with timely applications in biomedicine: (i) multi-view learning with paired samples, (ii) one covariate is time with unknown alignment, and (iii) multi-view learning without paired samples.

EPrint Type:Conference or Workshop Item (Paper)
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Learning/Statistics & Optimisation
Multimodal Integration
ID Code:7903
Deposited By:Ilkka Huopaniemi
Deposited On:17 March 2011