Key Instance Detection in Multi-Instance Learning

G. Liu; J. Wu; Z.-H. Zhou

Key Instance Detection in Multi-Instance Learning

G. Liu, J. Wu, Z.-H. Zhou ;

Proceedings of the Asian Conference on Machine Learning, PMLR 25:253-268, 2012.

Abstract

The goal of traditional multi-instance learning (MIL) is to predict the labels of the bags, whereas in many real applications, it is desirable to get the instance labels, especially the labels of key instances that trigger the bag labels, in addition to getting bag labels. Such a problem has been largely unexplored before. In this paper, we formulate the Key Instance Detection (KID) problem, and propose a voting framework (VF) solution to KID. The key of VF is to exploit the relationship among instances, represented by a citer kNN graph. This graph is different from commonly used nearest neighbor graphs, but is suitable for KID. Experiments validate the effectiveness of VF for KID. Additionally, VF also outperforms state-of-the-art MIL approaches on the performance of bag label prediction.

Related Material

Download PDF

@InProceedings{pmlr-v25-liu12b, title = {Key Instance Detection in Multi-Instance Learning}, author = {G. Liu and J. Wu and Z.-H. Zhou}, booktitle = {Proceedings of the Asian Conference on Machine Learning}, pages = {253--268}, year = {2012}, editor = {Steven C. H. Hoi and Wray Buntine}, volume = {25}, series = {Proceedings of Machine Learning Research}, address = {Singapore Management University, Singapore}, month = {04--06 Nov}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v25/liu12b/liu12b.pdf}, url = {http://proceedings.mlr.press/v25/liu12b.html}, abstract = {The goal of traditional multi-instance learning (MIL) is to predict the labels of the bags, whereas in many real applications, it is desirable to get the instance labels, especially the labels of key instances that trigger the bag labels, in addition to getting bag labels. Such a problem has been largely unexplored before. In this paper, we formulate the Key Instance Detection (KID) problem, and propose a voting framework (VF) solution to KID. The key of VF is to exploit the relationship among instances, represented by a citer kNN graph. This graph is different from commonly used nearest neighbor graphs, but is suitable for KID. Experiments validate the effectiveness of VF for KID. Additionally, VF also outperforms state-of-the-art MIL approaches on the performance of bag label prediction.} }

%0 Conference Paper %T Key Instance Detection in Multi-Instance Learning %A G. Liu %A J. Wu %A Z.-H. Zhou %B Proceedings of the Asian Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2012 %E Steven C. H. Hoi %E Wray Buntine %F pmlr-v25-liu12b %I PMLR %J Proceedings of Machine Learning Research %P 253--268 %U http://proceedings.mlr.press %V 25 %W PMLR %X The goal of traditional multi-instance learning (MIL) is to predict the labels of the bags, whereas in many real applications, it is desirable to get the instance labels, especially the labels of key instances that trigger the bag labels, in addition to getting bag labels. Such a problem has been largely unexplored before. In this paper, we formulate the Key Instance Detection (KID) problem, and propose a voting framework (VF) solution to KID. The key of VF is to exploit the relationship among instances, represented by a citer kNN graph. This graph is different from commonly used nearest neighbor graphs, but is suitable for KID. Experiments validate the effectiveness of VF for KID. Additionally, VF also outperforms state-of-the-art MIL approaches on the performance of bag label prediction.

TY - CPAPER TI - Key Instance Detection in Multi-Instance Learning AU - G. Liu AU - J. Wu AU - Z.-H. Zhou BT - Proceedings of the Asian Conference on Machine Learning PY - 2012/11/17 DA - 2012/11/17 ED - Steven C. H. Hoi ED - Wray Buntine ID - pmlr-v25-liu12b PB - PMLR SP - 253 DP - PMLR EP - 268 L1 - http://proceedings.mlr.press/v25/liu12b/liu12b.pdf UR - http://proceedings.mlr.press/v25/liu12b.html AB - The goal of traditional multi-instance learning (MIL) is to predict the labels of the bags, whereas in many real applications, it is desirable to get the instance labels, especially the labels of key instances that trigger the bag labels, in addition to getting bag labels. Such a problem has been largely unexplored before. In this paper, we formulate the Key Instance Detection (KID) problem, and propose a voting framework (VF) solution to KID. The key of VF is to exploit the relationship among instances, represented by a citer kNN graph. This graph is different from commonly used nearest neighbor graphs, but is suitable for KID. Experiments validate the effectiveness of VF for KID. Additionally, VF also outperforms state-of-the-art MIL approaches on the performance of bag label prediction. ER -