Metric learning

Abstract

Similarity metric learning is a sub-field of machine learning domain that focuses on developing techniques to measure the similarity between data points in a meaningful way. This project aims to design and develop a new similarity metric. The motivation for creating a new similarity function came from studying the limitations of existing similarity metrics, especially cosine similarity. The potential domains in machine learning that we identified where our similarity metric can be used are face recognition, face verification, image retrieval, etc.