Loss Functions in Clustering
Hadi Sadoghi Yazdi - Professor, Ferdowsi University of Mashhad
Thu, 19-May-2022 / 18:00 / Link:
Video Slides Poster


This presentation is about clustering with a brief overview of data visualization from the perspective of minimizing the loss function of data. In this talk, first, basic issues of clustering are discussed, by which we intend to communicate the fact that considering loss values is an essential part of the procedure, where in various problems, and also in relation to data types, we emphasize that the choice of loss functions is an influential factor in modeling the corresponding related optimization problems. In this presentation, we will address the impact of this factor in various applications. To clarify this, different examples are presented, such as clustering in social media, image tagging, noisy label detection, multi-view data fusion, and tracking. We present an example, in which the index of a data set is determined using different loss functions as well as a new fuzzy clustering criterion.


Hadi Sadoghi Yazdi received his B.S. degree in Electrical Engineering from Ferdowsi University of Mashhad in 1994, and then he received the M.S. and Ph.D. degrees in Electrical Engineering from Tarbiat Modares University in 1996 and 2005, respectively. Dr. Sadoghi Yazdi is currently a professor in the Department of Computer Engineering at Ferdowsi University of Mashhad, and his research interests cover pattern recognition, machine learning, machine vision, and signal processing. He has contributed to the field in a variety of topics and his research group is presently working on stock market signal processing.