Deterministic Annealing for Multiple-Instance Learning

Peter V. Gehler, Olivier Chapelle
; Proceedings of the Eleventh International Conference on Artificial Intelligence and Statistics, PMLR 2:123-130, 2007.

Abstract

In this paper we demonstrate how deterministic annealing can be applied to different SVM formulations of the multiple-instance learning (MIL) problem. Our results show that we find better local minima compared to the heuristic methods those problems are usually solved with. However this does not always translate into a better test error suggesting an inadequacy of the objective function. Based on this finding we propose a new objective function which together with the deterministic annealing algorithm finds better local minima and achieves better performance on a set of benchmark datasets. Furthermore the results also show how the structure of MIL datasets influence the performance of MIL algorithms and we discuss how future benchmark datasets for the MIL problem should be designed.

Cite this Paper


BibTeX
@InProceedings{pmlr-v2-gehler07a, title = {Deterministic Annealing for Multiple-Instance Learning}, author = {Peter V. Gehler and Olivier Chapelle}, booktitle = {Proceedings of the Eleventh International Conference on Artificial Intelligence and Statistics}, pages = {123--130}, year = {2007}, editor = {Marina Meila and Xiaotong Shen}, volume = {2}, series = {Proceedings of Machine Learning Research}, address = {San Juan, Puerto Rico}, month = {21--24 Mar}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v2/gehler07a/gehler07a.pdf}, url = {http://proceedings.mlr.press/v2/gehler07a.html}, abstract = {In this paper we demonstrate how deterministic annealing can be applied to different SVM formulations of the multiple-instance learning (MIL) problem. Our results show that we find better local minima compared to the heuristic methods those problems are usually solved with. However this does not always translate into a better test error suggesting an inadequacy of the objective function. Based on this finding we propose a new objective function which together with the deterministic annealing algorithm finds better local minima and achieves better performance on a set of benchmark datasets. Furthermore the results also show how the structure of MIL datasets influence the performance of MIL algorithms and we discuss how future benchmark datasets for the MIL problem should be designed.} }
Endnote
%0 Conference Paper %T Deterministic Annealing for Multiple-Instance Learning %A Peter V. Gehler %A Olivier Chapelle %B Proceedings of the Eleventh International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2007 %E Marina Meila %E Xiaotong Shen %F pmlr-v2-gehler07a %I PMLR %J Proceedings of Machine Learning Research %P 123--130 %U http://proceedings.mlr.press %V 2 %W PMLR %X In this paper we demonstrate how deterministic annealing can be applied to different SVM formulations of the multiple-instance learning (MIL) problem. Our results show that we find better local minima compared to the heuristic methods those problems are usually solved with. However this does not always translate into a better test error suggesting an inadequacy of the objective function. Based on this finding we propose a new objective function which together with the deterministic annealing algorithm finds better local minima and achieves better performance on a set of benchmark datasets. Furthermore the results also show how the structure of MIL datasets influence the performance of MIL algorithms and we discuss how future benchmark datasets for the MIL problem should be designed.
RIS
TY - CPAPER TI - Deterministic Annealing for Multiple-Instance Learning AU - Peter V. Gehler AU - Olivier Chapelle BT - Proceedings of the Eleventh International Conference on Artificial Intelligence and Statistics PY - 2007/03/11 DA - 2007/03/11 ED - Marina Meila ED - Xiaotong Shen ID - pmlr-v2-gehler07a PB - PMLR SP - 123 DP - PMLR EP - 130 L1 - http://proceedings.mlr.press/v2/gehler07a/gehler07a.pdf UR - http://proceedings.mlr.press/v2/gehler07a.html AB - In this paper we demonstrate how deterministic annealing can be applied to different SVM formulations of the multiple-instance learning (MIL) problem. Our results show that we find better local minima compared to the heuristic methods those problems are usually solved with. However this does not always translate into a better test error suggesting an inadequacy of the objective function. Based on this finding we propose a new objective function which together with the deterministic annealing algorithm finds better local minima and achieves better performance on a set of benchmark datasets. Furthermore the results also show how the structure of MIL datasets influence the performance of MIL algorithms and we discuss how future benchmark datasets for the MIL problem should be designed. ER -
APA
Gehler, P.V. & Chapelle, O.. (2007). Deterministic Annealing for Multiple-Instance Learning. Proceedings of the Eleventh International Conference on Artificial Intelligence and Statistics, in PMLR 2:123-130

Related Material