A review of multi-instance learning assumptions

James Foulds; Eibe Frank; James Foulds; Eibe Frank

doi:10.1017/S026988890999035X

Abstract: Multi-instance (MI) learning is a variant of inductive machine learning, where each learning example contains a bag of instances instead of a single feature vector. The term commonly refers to the supervised setting, where each bag is associated with a label. This type of representation is a natural fit for a number of real-world learning scenarios, including drug activity prediction and image classification, hence many MI learning algorithms have been proposed. Any MI learning method must relate instances to bag-level class labels, but many types of relationships between instances and class labels are possible. Although all early work in MI learning assumes a specific MI concept class known to be appropriate for a drug activity prediction domain; this ‘standard MI assumption’ is not guaranteed to hold in other domains. Much of the recent work in MI learning has concentrated on a relaxed view of the MI problem, where the standard MI assumption is dropped, and alternative assumptions are considered instead. However, often it is not clearly stated what particular assumption is used and how it relates to other assumptions that have been proposed. In this paper, we aim to clarify the use of alternative MI assumptions by reviewing the work done in this area.

Other Articles By Authors

A review of multi-instance learning assumptions

University of Waikato, Private Bag 3105, Hamilton

Published online: 01 March 2010

Abstract: Abstract: Multi-instance (MI) learning is a variant of inductive machine learning, where each learning example contains a bag of instances instead of a single feature vector. The term commonly refers to the supervised setting, where each bag is associated with a label. This type of representation is a natural fit for a number of real-world learning scenarios, including drug activity prediction and image classification, hence many MI learning algorithms have been proposed. Any MI learning method must relate instances to bag-level class labels, but many types of relationships between instances and class labels are possible. Although all early work in MI learning assumes a specific MI concept class known to be appropriate for a drug activity prediction domain; this ‘standard MI assumption’ is not guaranteed to hold in other domains. Much of the recent work in MI learning has concentrated on a relaxed view of the MI problem, where the standard MI assumption is dropped, and alternative assumptions are considered instead. However, often it is not clearly stated what particular assumption is used and how it relates to other assumptions that have been proposed. In this paper, we aim to clarify the use of alternative MI assumptions by reviewing the work done in this area.

HTML

In many cases, the instances are assumed to have hidden class labels that are in some way related to the labels for the bags. Depending on the problem domain, the prediction of the instance labels can also be an important task in its own right.

Microsoft Game Studios (1990).

Originally published in 2003 as a technical report at the University of Nebraska, Lincoln.

The medoid of a cluster is the element whose average distance to the other elements is minimal. In a geometric space, this is equivalent to choosing the element that is closest to the center of the cluster.

See also the later journal article (Dooly et al., 2002).

Not to be confused with Ray and Page’s (2001) primary instances, which are elements of a bag and are not assumed to ‘cause’ the other instances.

Rights and permissions

References (51)

About this article

Cite this article

James Foulds, Eibe Frank. 2010. A review of multi-instance learning assumptions. The Knowledge Engineering Review. 25: doi: 10.1017/S026988890999035X

James Foulds, Eibe Frank. 2010. A review of multi-instance learning assumptions. The Knowledge Engineering Review. 25: doi: 10.1017/S026988890999035X

{{lists.name}}

A review of multi-instance learning assumptions

Abstract

Rights and permissions

References

About this article

Cite this article

Article Metrics

Access History

Other Articles By Authors