Tuesday, September 26th, 2023
03:00pm - 04:00pm, in Wheatley 3-154-28

Maryam Bagherian

UMass Boston

Multi distance metric learning and its application

Abstract: Distance metric learning is an approach to exploring underlying structures within high-dimensional spaces. By learning a suitable distance metric, distance-based algorithms can better capture the intrinsic structure of data points, leading to improved performance. In contrast to single-metric learning approaches, multi-metric learning and geometric metric learning have demonstrated higher efficiency in handling complex data distributions and diverse data characteristics. These approaches offer increased flexibility and interpretability, making them particularly valuable for representation learning in complex multi-modal datasets. In this context, I provide a brief introduction to the concepts of metric learning and ideas for generalizing it to high-dimensional spaces and manifolds.