Parallel Filter-Based Feature Selection Based on Balanced Incomplete Block Designs

In this paper we propose a method for scaling up filterbased feature selection in classification problems. We use the conditional mutual information as filter measure and show how the required statistics can be computed in parallel avoiding unnecessary calculations. The distribution of the calculat...

Full description

Bibliographic Details
Main Authors: Salmerón Cerdán, Antonio, Madsen, Anders L., Jensen, Frank, Langseth, Helge, Nielsen, Thomas D., Ramos López, Darío, Martínez, Ana M., Masegosa, Andrés R.
Format: info:eu-repo/semantics/article
Language:English
Published: 2017
Subjects:
Online Access:http://hdl.handle.net/10835/4916
_version_ 1789406643980599296
author Salmerón Cerdán, Antonio
Madsen, Anders L.
Jensen, Frank
Langseth, Helge
Nielsen, Thomas D.
Ramos López, Darío
Martínez, Ana M.
Masegosa, Andrés R.
author_facet Salmerón Cerdán, Antonio
Madsen, Anders L.
Jensen, Frank
Langseth, Helge
Nielsen, Thomas D.
Ramos López, Darío
Martínez, Ana M.
Masegosa, Andrés R.
author_sort Salmerón Cerdán, Antonio
collection DSpace
description In this paper we propose a method for scaling up filterbased feature selection in classification problems. We use the conditional mutual information as filter measure and show how the required statistics can be computed in parallel avoiding unnecessary calculations. The distribution of the calculations between the available computing units is determined based on balanced incomplete block designs, a strategy first developed within the area of statistical design of experiments. We show the scalability of our method through a series of experiments on synthetic and real-world datasets.
format info:eu-repo/semantics/article
id oai:repositorio.ual.es:10835-4916
institution Universidad de Cuenca
language English
publishDate 2017
record_format dspace
spelling oai:repositorio.ual.es:10835-49162023-04-12T19:39:03Z Parallel Filter-Based Feature Selection Based on Balanced Incomplete Block Designs Salmerón Cerdán, Antonio Madsen, Anders L. Jensen, Frank Langseth, Helge Nielsen, Thomas D. Ramos López, Darío Martínez, Ana M. Masegosa, Andrés R. Bandpass filters Design of experiments Balanced incomplete block design Computing units Conditional mutual information Filter-based Real-world datasets Scaling-up Statistical design of experiments In this paper we propose a method for scaling up filterbased feature selection in classification problems. We use the conditional mutual information as filter measure and show how the required statistics can be computed in parallel avoiding unnecessary calculations. The distribution of the calculations between the available computing units is determined based on balanced incomplete block designs, a strategy first developed within the area of statistical design of experiments. We show the scalability of our method through a series of experiments on synthetic and real-world datasets. 2017-07-17T10:48:34Z 2017-07-17T10:48:34Z 2016 info:eu-repo/semantics/article The final publication is available at IOS Press through http://dx.doi.org/10.3233/978-1-61499-672-9-743 http://hdl.handle.net/10835/4916 10.3233/978-1-61499-672-9-743 en Attribution-NonCommercial-NoDerivatives 4.0 Internacional http://creativecommons.org/licenses/by-nc-nd/4.0/ info:eu-repo/semantics/openAccess
spellingShingle Bandpass filters
Design of experiments
Balanced incomplete block design
Computing units
Conditional mutual information
Filter-based
Real-world datasets
Scaling-up
Statistical design of experiments
Salmerón Cerdán, Antonio
Madsen, Anders L.
Jensen, Frank
Langseth, Helge
Nielsen, Thomas D.
Ramos López, Darío
Martínez, Ana M.
Masegosa, Andrés R.
Parallel Filter-Based Feature Selection Based on Balanced Incomplete Block Designs
title Parallel Filter-Based Feature Selection Based on Balanced Incomplete Block Designs
title_full Parallel Filter-Based Feature Selection Based on Balanced Incomplete Block Designs
title_fullStr Parallel Filter-Based Feature Selection Based on Balanced Incomplete Block Designs
title_full_unstemmed Parallel Filter-Based Feature Selection Based on Balanced Incomplete Block Designs
title_short Parallel Filter-Based Feature Selection Based on Balanced Incomplete Block Designs
title_sort parallel filter-based feature selection based on balanced incomplete block designs
topic Bandpass filters
Design of experiments
Balanced incomplete block design
Computing units
Conditional mutual information
Filter-based
Real-world datasets
Scaling-up
Statistical design of experiments
url http://hdl.handle.net/10835/4916
work_keys_str_mv AT salmeroncerdanantonio parallelfilterbasedfeatureselectionbasedonbalancedincompleteblockdesigns
AT madsenandersl parallelfilterbasedfeatureselectionbasedonbalancedincompleteblockdesigns
AT jensenfrank parallelfilterbasedfeatureselectionbasedonbalancedincompleteblockdesigns
AT langsethhelge parallelfilterbasedfeatureselectionbasedonbalancedincompleteblockdesigns
AT nielsenthomasd parallelfilterbasedfeatureselectionbasedonbalancedincompleteblockdesigns
AT ramoslopezdario parallelfilterbasedfeatureselectionbasedonbalancedincompleteblockdesigns
AT martinezanam parallelfilterbasedfeatureselectionbasedonbalancedincompleteblockdesigns
AT masegosaandresr parallelfilterbasedfeatureselectionbasedonbalancedincompleteblockdesigns