Recent non-linear feature selection approaches employing greedy optimisation
of Centred Kernel Target Alignment(KTA) exhibit strong results in terms of
generalisation accuracy and sparsity. However, they are computationally
prohibitive for large datasets. We propose randSel, a randomised feature
selection algorithm, with attractive scaling properties. Our theoretical
analysis of randSel provides strong probabilistic guarantees for correct
identification of relevant features. RandSel's characteristics make it an ideal
candidate for identifying informative learned representations. We've conducted
experimentation to establish the performance of this approach, and present
encouraging results, including a 3rd position result in the recent ICML black
box learning challenge as well as competitive results for signal peptide
prediction, an important problem in bioinformatics.
%0 Journal Article
%1 athanasakis2013principled
%A Athanasakis, Dimitrios
%A Shawe-Taylor, John
%A Fernandez-Reyes, Delmiro
%D 2013
%K feature-selection variable-selection
%T Principled Non-Linear Feature Selection
%U http://arxiv.org/abs/1312.5869
%X Recent non-linear feature selection approaches employing greedy optimisation
of Centred Kernel Target Alignment(KTA) exhibit strong results in terms of
generalisation accuracy and sparsity. However, they are computationally
prohibitive for large datasets. We propose randSel, a randomised feature
selection algorithm, with attractive scaling properties. Our theoretical
analysis of randSel provides strong probabilistic guarantees for correct
identification of relevant features. RandSel's characteristics make it an ideal
candidate for identifying informative learned representations. We've conducted
experimentation to establish the performance of this approach, and present
encouraging results, including a 3rd position result in the recent ICML black
box learning challenge as well as competitive results for signal peptide
prediction, an important problem in bioinformatics.
@article{athanasakis2013principled,
abstract = {Recent non-linear feature selection approaches employing greedy optimisation
of Centred Kernel Target Alignment(KTA) exhibit strong results in terms of
generalisation accuracy and sparsity. However, they are computationally
prohibitive for large datasets. We propose randSel, a randomised feature
selection algorithm, with attractive scaling properties. Our theoretical
analysis of randSel provides strong probabilistic guarantees for correct
identification of relevant features. RandSel's characteristics make it an ideal
candidate for identifying informative learned representations. We've conducted
experimentation to establish the performance of this approach, and present
encouraging results, including a 3rd position result in the recent ICML black
box learning challenge as well as competitive results for signal peptide
prediction, an important problem in bioinformatics.},
added-at = {2019-09-17T16:38:52.000+0200},
author = {Athanasakis, Dimitrios and Shawe-Taylor, John and Fernandez-Reyes, Delmiro},
biburl = {https://www.bibsonomy.org/bibtex/2a35d1ce23b659e55a363991561dde486/kirk86},
description = {[1312.5869v2] Principled Non-Linear Feature Selection},
interhash = {69d8c780b69ce7bdfcfb8a5f016e5dc1},
intrahash = {a35d1ce23b659e55a363991561dde486},
keywords = {feature-selection variable-selection},
note = {cite arxiv:1312.5869Comment: arXiv admin note: substantial text overlap with arXiv:1311.5636},
timestamp = {2019-09-17T16:38:52.000+0200},
title = {Principled Non-Linear Feature Selection},
url = {http://arxiv.org/abs/1312.5869},
year = 2013
}