Abstract
While the objective of the standard supervised learning problem is to classify feature vectors, in the multiple instance learning problem, the objective is to classify bags, where each bag contains multiple feature vectors. This represents a generalization of the standard problem, and this generalization becomes necessary in many real applications such as drug activity prediction, content-based image retrieval, and others. While the existing paradigms are based on learning the discriminant information either at the instance level or at the bag level, we propose to incorporate both levels of information. This is done by defining a discriminative embedding of the original space based on the responses of cluster-adapted instance classifiers. Results clearly show the advantage of the proposed method over the state of the art, where we tested the performance through a variety of well-known databases that come from real problems, and we also included an analysis of the performance using synthetically generated data.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Furthermore, in practice, the repetition of elements occurs very rarely. This is due to the fact that the elements \(\vec {x}_j\) are almost always defined as vectors of real components, i.e., \(\vec {x}_j \in \mathbb R^d\) for \(j = 1,\ldots ,N.\)
Given a multi-class problem, it can be reduced to several binary classification problems by means of common strategies such as one-against-all.
In order to simplify the discussion, we consider only clusters of positive instances in this explanation. However, the same idea can be used in clusters of negative instances. In this case, a cluster is contaminated if not only contains negative instances but it also contains many positive ones.
Regarding the negative instances, they were generated by using a constant number of Gaussian components, \(N_n=32\). We used a large number of components for the negative class in order to obtain realistic data. In this sense, note that the negative class is usually defined as the negation of the positive class, i.e., it groups all the classes of objects that are not positive, and thus forms an heterogeneous class.
References
Dietterich TG, Lathrop RH, Lozano-Perez T (1997) Solving the multiple-instance problem with axis-parallel rectangles. Artif Intell 89:31–71
Kotsiantis S, Kanellopoulos D (2008) Multi-instance learning for bankruptcy prediction. In: Proceedings of international conference on convergence and hybrid information technology, pp 1007–1012
Chen Y, Bi J, Wang JZ (2006) MILES: multiple-instance learning via embedded instance selection. IEEE Trans Pattern Anal Mach Intell 28:1931–1947
Reynolds DA, Quatieri TF, Dunn RB (2000) Speaker verification using adapted Gaussian mixture models. Dig Signal Process 10:19–41
Oded M (1998) Learning from ambiguity. PhD thesis, MIT
Zhang Q, Goldman SA (2001) EM-DD: an improved multiple-instance learning technique. In: Proceedings of neural information processing systems, pp 1073–1080
Andrews S, Tsochantaridis I, Hofmann T (2003) Support vector machines for multiple-instance learning. In: Proceedings of neural information processing systems, pp 561–568
Xu X, Frank E (2004) Logistic regression and boosting for labeled bags of instances. In: Proceedings of Pacific Asia conference on knowledge discovery and data mining, pp 272–281
Gehler PV, Chapelle O (2007) Deterministic annealing for multiple-instance learning. In: Proceedings of international conference on artificial intelligence and statistics, pp 123–130
Antic B, Ommer B (2013) Robust multiple-instance learning with superbags. Lecture notes in computer science, vol 7725, pp 242–255
Scott S, Zhang J, Brown J (2005) On generalized multiple-instance learning. Int J Comput Intell Appl 5:21–35
Zhou Z-H, Zhang M-L (2007) Solving multi-instance problems with classifier ensemble based on constructive clustering. Knowl Inf Syst 11:155–170
Gärtner T, Flach PA, Kowalczyk A, Smola AJ (2002) Multi-instance kernels. In: Proceedings of international conference on machine learning, pp 179–186
Foulds J, Frank E (2010) A review of multi-instance learning assumptions. Knowl Eng Rev 25:1–25
Nowak E, Jurie F, Triggs B (2006) Sampling strategies for bag-of-features image classification. In: Proceedings of European conference on computer vision, pp 490–503
Mikolajczyk K, Tuytelaars T, Schmid C, Zisserman A, Matas J, Schaffalitzky F, Kadir T, Gool LV (2005) A comparison of affine region detectors. Int J Comput Vis 65:43–72
Maron O, Lozano-Pérez T (1998) A framework for multiple-instance learning. In: Proceedings of neural information processing systems, pp 570–576
Bunescu R, Mooney R (2007) Multiple instance learning for sparse positive bags. In: Proceedings of international conference on machine learning, pp 105–112
Zhou Z-H, Sun Y-Y, Li Y-F (2009), Multi-instance learning by treating instances as non i.i.d. samples. In: Proceedings of international conference on machine learning, pp 1249–1256
Zhang M-L, Zhou Z-H (2009) Multi-instance clustering with applications to multi-instance prediction. Appl Intell 31:47–68
Fan R-E, Chang K-W, Hsieh C-J, Wang X-R, Lin C-J (2008) Liblinear: a library for large linear classification. J Mach Learn Res 9:1871–1874
Hsu C-W, Chang C-C, Lin C-J (2003) A practical guide to support vector classification. Technical report, National Taiwan University
Yang J (2005) Mill: a multi-instance learning library. http://www-2.cs.cmu.edu/juny/IR-lab/
Acknowledgments
We thank anonymous reviewers for their very useful comments and suggestions. This work was supported by the fellowship RYC-2008-03789 and the Spanish project TRA2011-29454-C03-01.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Amores, J. MILDE: multiple instance learning by discriminative embedding. Knowl Inf Syst 42, 381–407 (2015). https://doi.org/10.1007/s10115-013-0711-1
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10115-013-0711-1