Multi-task metric learning was introduced by Caruana in 1997. The performance is improved by considering multiple learning tasks and sharing information with other tasks. The metric is used as a measure of similarity or dissimilarity and there are various distance metrics such as Euclidean distance, cosine similarity, Hamming distance, etc. there are various evolving challenges in obtaining training data set which has become a costly process. To overcome these problems multi-task has to be introduced. This article includes the basic concepts, strategies and applications of metric learning.

To learn and know more please refer this link:

https://link.springer.com/article/10.1186/s41044-018-0029-9