Label Distribution Learning
Abstract
Multi-label is the natural attribute of things. Describing instances with one label could lead to massive loss of information. Instead of assigning a single label, attaching multiple labels to instances suffers from label ambiguity. That is because the weight of each label is different. Label distribution is introduced to adapt to the situation.
Date
May 3, 2017
9:00 AM
— 10:00 AM
Event
Lab Seminar
Location
Room 311, Innovation Center, UESTC
Wei Han
Ph.D. Candidate
My research interests include interpretable machine learning, transfer Learning, continual / lifelong learning, few-shot learning and adversarial learning.