Seminar by Dr. Marcus de Carvalho from Post-Doc, CCDS; Research Manager, SIMTech/A*STAR, 15Aug 2024, CIL Meeting Room

Title     : Towards Cross-Domain Continual Learning
Date    : 15th August, 2024
Time    : 11.30 AM To 12.00 PM
Venue : CIL Meeting Room (N4-B1A-02)
Presenter: Dr. Marcus de Carvalho

Abstract:

Continual Learning is a process that involves training learning agents to sequentially master a stream of tasks or classes without revisiting past data. The challenge lies in leveraging previously acquired knowledge to learn new tasks efficiently while avoiding catastrophic forgetting. Existing methods primarily focus on single domains, restricting their applicability to specific problems. In this talk, I will introduce a novel approach called Cross-Domain Continual Learning (CDCL) that addresses the limitation of being constrained to single supervised domains. Our method combines inter- and intra-task cross-attention mechanisms within a compact convolutional network. Finally, I will also share some of the strategies that motivated our approach and some techniques that may be useful for enhancing it and applying it to other fields.

Biography:

Marcus has a BEng in Computer Engineering from Centro Universitário de Brasília (CEUB) - Brazil, an MSc in electrical engineering and computer science from the National Chiao Tung University (NCTU) - Taiwan, and a PhD in computer science and engineering from Nanyang Technological University (NTU) - Singapore, where most of his research involved the fields of Evolutionary Algorithms, Deep Transfer Learning, Lifelong Learning and Online Learning, applied on Computer Vision, Data Streaming, Data Compression and Autonomous Agents. He is currently a Post-Doctorate supervised by Prof. Zhang Jie at the College of Computing and Data Science at NTU and a Research Manager at SIMTech/A*STAR. Additionally, previously Marcus was the CTO and Co-Founder of Pin People from 2016 to 2018, an HRTech Startup that has been awarded 3 times on the list of top 100 startups in Brazil. Alongside his research, Marcus is also interested in Computer Graphics, Low-Level Programming and Emulators.