My SciELO
Services on Demand
Journal
Article
Indicators
- Cited by SciELO
Related links
- Similars in SciELO
Share
Revista Cubana de Ciencias Informáticas
On-line version ISSN 2227-1899
Abstract
PEREZ VERONA, Isabel Cristina and ARCO GARCIA, Leticia. A brief review on unsupervised metric learning. Rev cuba cienc informat [online]. 2016, vol.10, n.4, pp.43-67. ISSN 2227-1899.
ABSTRACT Several machine learning methods rely on the notion of distances in a multidimensional space, these distances are used for estimating the similarity between two objects, according to historical data. In such cases, when the metric is specifically designed to the context, better results are often obtained. However, designing a metric is a complex task. Metric learning automatically learns a distance metric according to the characteristics of the data. Unsupervised metric learning algorithms have achieved good results in cases where there is not available much information about the data. These algorithms do not require class label information, they are applied to improve unsupervised machine learning methods, mainly for improving clustering results. Here we will mention some of the recent works done in this area, their advantages, disadvantages and applications.
Keywords : unsupervised metric learning; dimensionality reduction; distance.