A Probabilistic Model for Dirty Multi-task Feature Selection

Daniel Hernandez-Lobato, Jose Miguel Hernandez-Lobato, Zoubin Ghahramani
Proceedings of the 32nd International Conference on Machine Learning, PMLR 37:1073-1082, 2015.

Abstract

Multi-task feature selection methods often make the hypothesis that learning tasks share relevant and irrelevant features. However, this hypothesis may be too restrictive in practice. For example, there may be a few tasks with specific relevant and irrelevant features (outlier tasks). Similarly, a few of the features may be relevant for only some of the tasks (outlier features). To account for this, we propose a model for multi-task feature selection based on a robust prior distribution that introduces a set of binary latent variables to identify outlier tasks and outlier features. Expectation propagation can be used for efficient approximate inference under the proposed prior. Several experiments show that a model based on the new robust prior provides better predictive performance than other benchmark methods.

Cite this Paper

BibTeX
@InProceedings{pmlr-v37-hernandez-lobatoa15, title = {A Probabilistic Model for Dirty Multi-task Feature Selection}, author = {Hernandez-Lobato, Daniel and Hernandez-Lobato, Jose Miguel and Ghahramani, Zoubin}, booktitle = {Proceedings of the 32nd International Conference on Machine Learning}, pages = {1073--1082}, year = {2015}, editor = {Bach, Francis and Blei, David}, volume = {37}, series = {Proceedings of Machine Learning Research}, address = {Lille, France}, month = {07--09 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v37/hernandez-lobatoa15.pdf}, url = {https://proceedings.mlr.press/v37/hernandez-lobatoa15.html}, abstract = {Multi-task feature selection methods often make the hypothesis that learning tasks share relevant and irrelevant features. However, this hypothesis may be too restrictive in practice. For example, there may be a few tasks with specific relevant and irrelevant features (outlier tasks). Similarly, a few of the features may be relevant for only some of the tasks (outlier features). To account for this, we propose a model for multi-task feature selection based on a robust prior distribution that introduces a set of binary latent variables to identify outlier tasks and outlier features. Expectation propagation can be used for efficient approximate inference under the proposed prior. Several experiments show that a model based on the new robust prior provides better predictive performance than other benchmark methods.} }
Endnote
%0 Conference Paper %T A Probabilistic Model for Dirty Multi-task Feature Selection %A Daniel Hernandez-Lobato %A Jose Miguel Hernandez-Lobato %A Zoubin Ghahramani %B Proceedings of the 32nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2015 %E Francis Bach %E David Blei %F pmlr-v37-hernandez-lobatoa15 %I PMLR %P 1073--1082 %U https://proceedings.mlr.press/v37/hernandez-lobatoa15.html %V 37 %X Multi-task feature selection methods often make the hypothesis that learning tasks share relevant and irrelevant features. However, this hypothesis may be too restrictive in practice. For example, there may be a few tasks with specific relevant and irrelevant features (outlier tasks). Similarly, a few of the features may be relevant for only some of the tasks (outlier features). To account for this, we propose a model for multi-task feature selection based on a robust prior distribution that introduces a set of binary latent variables to identify outlier tasks and outlier features. Expectation propagation can be used for efficient approximate inference under the proposed prior. Several experiments show that a model based on the new robust prior provides better predictive performance than other benchmark methods.
RIS
TY - CPAPER TI - A Probabilistic Model for Dirty Multi-task Feature Selection AU - Daniel Hernandez-Lobato AU - Jose Miguel Hernandez-Lobato AU - Zoubin Ghahramani BT - Proceedings of the 32nd International Conference on Machine Learning DA - 2015/06/01 ED - Francis Bach ED - David Blei ID - pmlr-v37-hernandez-lobatoa15 PB - PMLR DP - Proceedings of Machine Learning Research VL - 37 SP - 1073 EP - 1082 L1 - http://proceedings.mlr.press/v37/hernandez-lobatoa15.pdf UR - https://proceedings.mlr.press/v37/hernandez-lobatoa15.html AB - Multi-task feature selection methods often make the hypothesis that learning tasks share relevant and irrelevant features. However, this hypothesis may be too restrictive in practice. For example, there may be a few tasks with specific relevant and irrelevant features (outlier tasks). Similarly, a few of the features may be relevant for only some of the tasks (outlier features). To account for this, we propose a model for multi-task feature selection based on a robust prior distribution that introduces a set of binary latent variables to identify outlier tasks and outlier features. Expectation propagation can be used for efficient approximate inference under the proposed prior. Several experiments show that a model based on the new robust prior provides better predictive performance than other benchmark methods. ER -
APA
Hernandez-Lobato, D., Hernandez-Lobato, J.M. & Ghahramani, Z.. (2015). A Probabilistic Model for Dirty Multi-task Feature Selection. Proceedings of the 32nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 37:1073-1082 Available from https://proceedings.mlr.press/v37/hernandez-lobatoa15.html.

Related Material