Transferring Knowledge by Prior Feature Sampling

Victor Eruhimov, Vladimir Martyanov, Aleksey Polovinkin
Proceedings of the Workshop on New Challenges for Feature Selection in Data Mining and Knowledge Discovery at ECML/PKDD 2008, PMLR 4:135-147, 2008.

Abstract

The paper presents a novel method for transfer learning through prior variable sampling. A set of problems defined in the same feature space with similar dependencies of target on features is considered. We suggest a method for learning a decision tree ensemble on each of the problems by prior estimation of variable importance on other problems in the set and using it for regularizing model learning for a small amount of training samples. The method is tested on several simulated and real datasets. In particular, we apply our method for a set of time series classification (TSC) problems. Our analysis demonstrates an intriguing result: a model trained on several TSC problems can learn a new problem with high accuracy from a low number of samples.

Cite this Paper

BibTeX
@InProceedings{pmlr-v4-eruhimov08a, title = {Transferring Knowledge by Prior Feature Sampling}, author = {Eruhimov, Victor and Martyanov, Vladimir and Polovinkin, Aleksey}, booktitle = {Proceedings of the Workshop on New Challenges for Feature Selection in Data Mining and Knowledge Discovery at ECML/PKDD 2008}, pages = {135--147}, year = {2008}, editor = {Saeys, Yvan and Liu, Huan and Inza, Iñaki and Wehenkel, Louis and Pee, Yves Van de}, volume = {4}, series = {Proceedings of Machine Learning Research}, address = {Antwerp, Belgium}, month = {15 Sep}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v4/eruhimov08a/eruhimov08a.pdf}, url = {https://proceedings.mlr.press/v4/eruhimov08a.html}, abstract = {The paper presents a novel method for transfer learning through prior variable sampling. A set of problems defined in the same feature space with similar dependencies of target on features is considered. We suggest a method for learning a decision tree ensemble on each of the problems by prior estimation of variable importance on other problems in the set and using it for regularizing model learning for a small amount of training samples. The method is tested on several simulated and real datasets. In particular, we apply our method for a set of time series classification (TSC) problems. Our analysis demonstrates an intriguing result: a model trained on several TSC problems can learn a new problem with high accuracy from a low number of samples.} }
Endnote
%0 Conference Paper %T Transferring Knowledge by Prior Feature Sampling %A Victor Eruhimov %A Vladimir Martyanov %A Aleksey Polovinkin %B Proceedings of the Workshop on New Challenges for Feature Selection in Data Mining and Knowledge Discovery at ECML/PKDD 2008 %C Proceedings of Machine Learning Research %D 2008 %E Yvan Saeys %E Huan Liu %E Iñaki Inza %E Louis Wehenkel %E Yves Van de Pee %F pmlr-v4-eruhimov08a %I PMLR %P 135--147 %U https://proceedings.mlr.press/v4/eruhimov08a.html %V 4 %X The paper presents a novel method for transfer learning through prior variable sampling. A set of problems defined in the same feature space with similar dependencies of target on features is considered. We suggest a method for learning a decision tree ensemble on each of the problems by prior estimation of variable importance on other problems in the set and using it for regularizing model learning for a small amount of training samples. The method is tested on several simulated and real datasets. In particular, we apply our method for a set of time series classification (TSC) problems. Our analysis demonstrates an intriguing result: a model trained on several TSC problems can learn a new problem with high accuracy from a low number of samples.
RIS
TY - CPAPER TI - Transferring Knowledge by Prior Feature Sampling AU - Victor Eruhimov AU - Vladimir Martyanov AU - Aleksey Polovinkin BT - Proceedings of the Workshop on New Challenges for Feature Selection in Data Mining and Knowledge Discovery at ECML/PKDD 2008 DA - 2008/09/11 ED - Yvan Saeys ED - Huan Liu ED - Iñaki Inza ED - Louis Wehenkel ED - Yves Van de Pee ID - pmlr-v4-eruhimov08a PB - PMLR DP - Proceedings of Machine Learning Research VL - 4 SP - 135 EP - 147 L1 - http://proceedings.mlr.press/v4/eruhimov08a/eruhimov08a.pdf UR - https://proceedings.mlr.press/v4/eruhimov08a.html AB - The paper presents a novel method for transfer learning through prior variable sampling. A set of problems defined in the same feature space with similar dependencies of target on features is considered. We suggest a method for learning a decision tree ensemble on each of the problems by prior estimation of variable importance on other problems in the set and using it for regularizing model learning for a small amount of training samples. The method is tested on several simulated and real datasets. In particular, we apply our method for a set of time series classification (TSC) problems. Our analysis demonstrates an intriguing result: a model trained on several TSC problems can learn a new problem with high accuracy from a low number of samples. ER -
APA
Eruhimov, V., Martyanov, V. & Polovinkin, A.. (2008). Transferring Knowledge by Prior Feature Sampling. Proceedings of the Workshop on New Challenges for Feature Selection in Data Mining and Knowledge Discovery at ECML/PKDD 2008, in Proceedings of Machine Learning Research 4:135-147 Available from https://proceedings.mlr.press/v4/eruhimov08a.html.

Related Material