Mahardhika Pratama, University of South Australia, Australia

Tuesday, October 28, 2025
08:45 – 09:45 IRST

Title:Handling Dynamic Environments with Continual and Autonomous Machine Learning

Abstract

Although deep learning (DL) has achieved promising performances in many applications, its design principle lies in the i.i.d assumption where the learning environment is static. It does not adapt to changing learning environments without suffering from the catastrophic forgetting problem where it loses its performance to previous tasks when adapting to new tasks. In this talk, I will cover the continual learning problem including its solutions. It encompasses the problems of continual learning, few-shot learning, few-shot continual learning, cross-domain continual learning, and federated continual learning. 

Biography

A/Prof. Mahardhika Pratama is an associate-professor-level enterprise fellow at the academic unit of STEM, University of South Australia. He received his PhD from the University of New South Wales in 2014. His research interests are continual learning, few-shot learning, domain adaptation, and few-shot learning.