Learning neighborhoods for metric learning

Authors:
Jun Wang;Adam Woznica;Alexandros Kalousis
Affiliations:
AI Lab, Department of Computer Science, University of Geneva, Switzerland, Department of Business Informatics, University of Applied Sciences, Western Switzerland;AI Lab, Department of Computer Science, University of Geneva, Switzerland, Department of Business Informatics, University of Applied Sciences, Western Switzerland;AI Lab, Department of Computer Science, University of Geneva, Switzerland, Department of Business Informatics, University of Applied Sciences, Western Switzerland
Venue:
ECML PKDD'12 Proceedings of the 2012 European conference on Machine Learning and Knowledge Discovery in Databases - Volume Part I
Year:
2012

Citing 10
Cited 0

Theory of linear and integer programming

Theory of linear and integer programming
Some Notes on Alternating Optimization

AFSS '02 Proceedings of the 2002 AFSS International Conference on Fuzzy Systems. Calcutta: Advances in Soft Computing
Stability of feature selection algorithms: a study on high-dimensional spaces

Knowledge and Information Systems
Information-theoretic metric learning

Proceedings of the 24th international conference on Machine learning
A tutorial on spectral clustering

Statistics and Computing
Metric Learning: A Support Vector Approach

ECML PKDD '08 Proceedings of the European conference on Machine Learning and Knowledge Discovery in Databases - Part II
Graph construction and b-matching for semi-supervised learning

ICML '09 Proceedings of the 26th Annual International Conference on Machine Learning
Geometry-aware metric learning

ICML '09 Proceedings of the 26th Annual International Conference on Machine Learning
Distance Metric Learning for Large Margin Nearest Neighbor Classification

The Journal of Machine Learning Research
Regularized neighborhood component analysis

SCIA'07 Proceedings of the 15th Scandinavian conference on Image analysis

Quantified Score

Hi-index	0.00

Visualization

Abstract

Metric learning methods have been shown to perform well on different learning tasks. Many of them rely on target neighborhood relationships that are computed in the original feature space and remain fixed throughout learning. As a result, the learned metric reflects the original neighborhood relations. We propose a novel formulation of the metric learning problem in which, in addition to the metric, the target neighborhood relations are also learned in a two-step iterative approach. The new formulation can be seen as a generalization of many existing metric learning methods. The formulation includes a target neighbor assignment rule that assigns different numbers of neighbors to instances according to their quality; 'high quality' instances get more neighbors. We experiment with two of its instantiations that correspond to the metric learning algorithms LMNN and MCML and compare it to other metric learning methods on a number of datasets. The experimental results show state-of-the-art performance and provide evidence that learning the neighborhood relations does improve predictive performance.