Skip to Main content Skip to Navigation
Journal articles

Mixture of Gaussians for Distance Estimation with Missing Data

Emil Eirola 1, * Amaury Lendasse 2 Vincent Vandewalle 3, 4 Christophe Biernacki 3
* Corresponding author
3 MODAL - MOdel for Data Analysis and Learning
Inria Lille - Nord Europe, LPP - Laboratoire Paul Painlevé - UMR 8524, METRICS - Evaluation des technologies de santé et des pratiques médicales - ULR 2694, Polytech Lille - École polytechnique universitaire de Lille, Université de Lille, Sciences et Technologies
Abstract : The majority of all commonly used machine learning methods can not be applied directly to data sets with missing values. However, most such meth- ods only depend on the relative di erences between samples instead of their particular values, and thus one useful approach is to directly estimate the pairwise distances between all samples in the data set. This is accomplished by tting a Gaussian mixture model to the data, and using it to derive estimates for the distances. Experimental simulations con rm that the pro- posed method provides accurate estimates compared to alternative methods for estimating distances. The experimental evaluation additionally shows that more accurately estimating distances leads to improved prediction performance for classification and regression tasks when used as inputs for a neural network.
Document type :
Journal articles
Complete list of metadata

Cited literature [29 references]  Display  Hide  Download

https://hal.inria.fr/hal-00921023
Contributor : Christophe Biernacki <>
Submitted on : Tuesday, December 30, 2014 - 2:54:47 PM
Last modification on : Monday, January 18, 2021 - 2:08:56 PM
Long-term archiving on: : Tuesday, March 31, 2015 - 10:22:02 AM

File

Publication206.pdf
Files produced by the author(s)

Identifiers

Collections

Citation

Emil Eirola, Amaury Lendasse, Vincent Vandewalle, Christophe Biernacki. Mixture of Gaussians for Distance Estimation with Missing Data. Neurocomputing, Elsevier, 2014, 131, pp.32-42. ⟨10.1016/j.neucom.2013.07.050⟩. ⟨hal-00921023v2⟩

Share

Metrics

Record views

463

Files downloads

1415