Avatar

Thang Doan


News

April 2021 : Our workshop proposal "Theory of Continual Learning" for ICML 2021 has been accepted !

January 2021 : Paper accepted at AISTATS 2021 A Theoretical Analysis of Catastrophic Forgetting through the NTK Overlap Matrix

January 2021 : Paper accepted at ICLR 2021 Regularized Inverse Reinforcement Learning

December 2020 : Our workshop proposal "Self-Supervised learning for Reinforcement Learning" for ICLR 2021 has been accepted !

November 2020 : Prepint Generalisation guarantees for continual learning with orthogonal gradient descent

October 2020 : Paper accepted at Neurips 2020 Deep RL workshop Domain Adversarial Reinforcement Learning

September 2020 : Paper accepted at Neurips 2020 Deep Reinforcement and InfoMax Learning [Poster]


Bio

I am a postdoctoral researcher @ McGill University and Mila (Quebec AI Institute) under supervision of Professor Joelle Pineau. My interests span representation and reinforcement learning, generative models and continual learning. I dedicated a part of my research investigating robust representation methods for better generalization in reinforcement learning. My other stream of focus seeks better understanding of Catastrophic Forgetting and Inference.

I earned my PhD in Operations Management from McGill University, focusing my research in exploration problems for Deep Reinforcement Learning and Generative models (GANs). Prior to this, I completed a French mechanical engineering degree from Arts et Metiers ParisTech with a master degree (MSc) in Mathematical Modeling at University College London.


Interests

  • Artificial Intelligence
  • Reinforcement Learning
  • Generative models
  • Deep Learning
  • Continual Learning

Education

  • PhD in Operation Management, 2019

    McGill University, Canada

  • MSc in Mathematical Modeling, 2013

    University College London, UK

  • Mechanical Engineering Degree, 2013

    Arts et Metiers ParisTech, France