Largest Source Subset Selection for Instance Transfer

Shuang Zhou, Gijs Schoenmakers, Evgueni Smirnov, Ralf Peeters, Kurt Driessens, Siqi Chen
Asian Conference on Machine Learning, PMLR 45:423-438, 2016.

Abstract

Instance-transfer learning has emerged as a promising learning framework to boost performance of prediction models on newly-arrived tasks. The success of the framework depends on the relevance of the source data to the target data. This paper proposes a new approach to source data selection for instance-transfer learning. The approach is capable of selecting the largest subset S^* of the source data which relevance to the target data is statistically guaranteed to be the highest among any superset of S^*. The approach is formally described and theoretically justified. Experimental results on real-world data sets demonstrate that the approach outperforms existing instance selection methods.

Cite this Paper

BibTeX
@InProceedings{pmlr-v45-Zhou15, title = {Largest Source Subset Selection for Instance Transfer}, author = {Zhou, Shuang and Schoenmakers, Gijs and Smirnov, Evgueni and Peeters, Ralf and Driessens, Kurt and Chen, Siqi}, booktitle = {Asian Conference on Machine Learning}, pages = {423--438}, year = {2016}, editor = {Holmes, Geoffrey and Liu, Tie-Yan}, volume = {45}, series = {Proceedings of Machine Learning Research}, address = {Hong Kong}, month = {20--22 Nov}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v45/Zhou15.pdf}, url = {https://proceedings.mlr.press/v45/Zhou15.html}, abstract = {Instance-transfer learning has emerged as a promising learning framework to boost performance of prediction models on newly-arrived tasks. The success of the framework depends on the relevance of the source data to the target data. This paper proposes a new approach to source data selection for instance-transfer learning. The approach is capable of selecting the largest subset S^* of the source data which relevance to the target data is statistically guaranteed to be the highest among any superset of S^*. The approach is formally described and theoretically justified. Experimental results on real-world data sets demonstrate that the approach outperforms existing instance selection methods. } }
Endnote
%0 Conference Paper %T Largest Source Subset Selection for Instance Transfer %A Shuang Zhou %A Gijs Schoenmakers %A Evgueni Smirnov %A Ralf Peeters %A Kurt Driessens %A Siqi Chen %B Asian Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2016 %E Geoffrey Holmes %E Tie-Yan Liu %F pmlr-v45-Zhou15 %I PMLR %P 423--438 %U https://proceedings.mlr.press/v45/Zhou15.html %V 45 %X Instance-transfer learning has emerged as a promising learning framework to boost performance of prediction models on newly-arrived tasks. The success of the framework depends on the relevance of the source data to the target data. This paper proposes a new approach to source data selection for instance-transfer learning. The approach is capable of selecting the largest subset S^* of the source data which relevance to the target data is statistically guaranteed to be the highest among any superset of S^*. The approach is formally described and theoretically justified. Experimental results on real-world data sets demonstrate that the approach outperforms existing instance selection methods.
RIS
TY - CPAPER TI - Largest Source Subset Selection for Instance Transfer AU - Shuang Zhou AU - Gijs Schoenmakers AU - Evgueni Smirnov AU - Ralf Peeters AU - Kurt Driessens AU - Siqi Chen BT - Asian Conference on Machine Learning DA - 2016/02/25 ED - Geoffrey Holmes ED - Tie-Yan Liu ID - pmlr-v45-Zhou15 PB - PMLR DP - Proceedings of Machine Learning Research VL - 45 SP - 423 EP - 438 L1 - http://proceedings.mlr.press/v45/Zhou15.pdf UR - https://proceedings.mlr.press/v45/Zhou15.html AB - Instance-transfer learning has emerged as a promising learning framework to boost performance of prediction models on newly-arrived tasks. The success of the framework depends on the relevance of the source data to the target data. This paper proposes a new approach to source data selection for instance-transfer learning. The approach is capable of selecting the largest subset S^* of the source data which relevance to the target data is statistically guaranteed to be the highest among any superset of S^*. The approach is formally described and theoretically justified. Experimental results on real-world data sets demonstrate that the approach outperforms existing instance selection methods. ER -
APA
Zhou, S., Schoenmakers, G., Smirnov, E., Peeters, R., Driessens, K. & Chen, S.. (2016). Largest Source Subset Selection for Instance Transfer. Asian Conference on Machine Learning, in Proceedings of Machine Learning Research 45:423-438 Available from https://proceedings.mlr.press/v45/Zhou15.html.

Related Material