Multiple instance learning via distance metric optimization

Publication Type:
Conference Proceeding
IEEE International Conference on Image Processing, ICIP 2013, 2013, pp. 2617 - 2621
Issue Date:
Full metadata record
Files in This Item:
Filename Description Size
Thumbnail2013003393OK.pdf671.9 kB
Adobe PDF
Multiple Instance Learning (MIL) has been widely applied in practice, such as drug activity prediction, content-based im- age retrieval. In MIL, a sample, comprised of a set of in- stances, is called a bag. Labels are assigned to bags instead of instances. The uncertainty of labels on instances makes MIL different from conventional supervised single instance learn- ing (SIL) tasks. Therefore, it is critical to learn an effective mapping to convert an MIL task to an SIL task. In this pa- per, we present OptMILES by learning the optimal transfor- mation on the bag-to-instance similarity measure, exploring the optimal distance metric between instances, by an alternat- ing minimization training procedure. We thoroughly evalu- ate the proposed method on both a synthetic dataset and real world datasets by comparing with representative MIL algo- rithms. The experimental results suggest the effectiveness of OptMILES
Please use this identifier to cite or link to this item: