Riashat Islam
riashat dot islam dot 93 at gmail dot com

I am a Research Scientist in Machine Learning (Reinforcement Learning and Generative AI). Previously, I was a PhD student supervised by Doina Precup in the Reasoning and Learning Lab at McGill University and a part of Mila - Quebec AI Institute. I am currently a Student Researcher at Microsoft Research NYC and Microsoft Research Montreal, where I work with John Langford, Alex Lamb, Dipendra Misra, Remi Tachet Des Combes, Romain Laroche and Harm Van Seijen. Previously, I was also a research intern at Microsoft Research Montreal, working with Phil Bachman. I also collaborate closely with Anirudh Goyal and Yoshua Bengio. For past and present collaborators, please see the list here.

I completed my Masters at University of Cambridge in the MPhil Machine Learning, Speech and Language Technology Program, under the supervision of Zoubin Ghahramani and Yarin Gal in the Cambridge Machine Learning Group. My Masters was funded by the Cambridge Commonwealth and International Trust, and I was a member of St John's College, Cambridge. 

Prior to that I studied Electronic and Electrical Engineering at University College London working under supervision of John Shawe-Taylor and Guy Lever from the Centre of Computational Statistics and Machine Learning and Gatsby Computational and Neuroscience Unit at UCL. I also had a great fortune working with David Silver from Google DeepMind during my undergraduate thesis.

I was also a summer research student at California Institute of Technology (Caltech) in the Summer Undergraduate Research Fellowship program (SURF) where I worked under supervision of Richard Murray in the Control and Dynamical Systems Lab, under a project in collaboration with the NASA Jet Propulsion Laboratory (JPL), and Keck Institute for Space Studies.  Before that, I was a summer research student in the Machine Learning Group at Johns Hopkins University working under supervision of Suchi Saria.

Google Scholar  /  GitHub  /  DBLP  /  Twitter  /  LinkedIn

Academic Places
  • (2017 - 2023) PhD, McGill University (Supervisor : Doina Precup)
  • (2016 - 2017) MPhil, University of Cambridge (Supervisor : Zoubin Ghahramani)
  • (2011 - 2015) Undergraduate, University College London, UCL (Supervisor : John Shawe-Taylor)
Research Work Experience
  • (Sep'23 - Present) Research Scientist, DreamFold AI
  • (Jun'23 - Oct'23) Research Intern, Morgan Stanley, NYC
  • (Jan'22 - Apr'23) Student Researcher, Microsoft Research (Advisor : J. Langford, H. Seijen, R.T.D Combes)
  • (Jun'21 - Nov'21) Research Intern, Microsoft Research NYC (Advisor : John Langford)
  • (Jan'18 - Feb'19) Student Researcher, Microsoft Research Montreal (Advisor : Phil Bachman)
  • (Oct'17 - Dec'17) Research Intern, Microsoft Research Montreal, Maluuba (Advisor : Phil Bachman)
  • (Jun'15 - Oct'15) Research Intern, Caltech Surf Program (Advisor : Richard Murray)
  • (Jun'14 - Oct'14) Research Intern, Johns Hopkins SRE Program (Advisor : Suchi Saria)
  • (Jun'13 - May'14) Industrial Placement Intern, J.P. Morgan
  • (Jun'12 - Oct'12) Intern, Deutsche Bank

Learning Latent Dynamic Robust Representations for World Models
Ruixiang Sun, Hongyu Zang, Xin Li, Riashat Islam
ICML 2024

PcLast : Discovering Plannable Continuous Latent States
Anurag Koul, Shivakanth Sujit, Shaoru Chen, Ben Evans, Lili Wu, Riashat Islam, Raihan Seraj, Yonathan Efroni, Miroslav Dudik, John Langford, Alex Lamb
ICML 2024

Generalizing Inverse Kinematics to Discover Agent-Centric Dynamics for Finite-Memory POMDPs
Lili Wu, Ben Evans, Riashat Islam, Raihan Seraj, Yonathan Efroni, Alex Lamb
In Submission, ArXiv

EQA-MX: Embodied Question Answering using Multimodal Expression
Md Mofijul Islam, Alexi Gladstone, Riashat Islam, Tariq Iqbal

Ignorance is Bliss: Robust Control via Information Gating
Manan Tomar, Riashat Islam, Matthew E. Taylor, Sergey Levine, Philip Bachman
NeurIPS 2023

Understanding and Addressing the Pitfalls of Bisimulation-based Representations in Offline Reinforcement Learning
Hongyu Zang, Xin Li, Leiji Zhang, Yang Liu, Baigui Sun, Riashat Islam, Remi Tachet des Combes, Romain Laroche
NeurIPS 2023

Agent-Controller Representations : Principled Offline RL with Rich Exogenous Information
Riashat Islam, Manan Tomar, Alex Lamb, Yonathan Efroni, Hongyu Zang, Aniket Didolkar, Dipendra Misra, Xin Li, Harm Van Seijen, Remi Tachet Des Combes, John Langford.
ICML 2023

Discrete Factorial Representations as an Abstraction for Goal Conditioned RL
Riashat Islam, Hongyu Zang, Anirudh Goyal, Alex Lamb, Kenji Kawaguchi, Xin Li, Romain Laroche, Yoshua Bengio, Remi Tachet Des Combes
NeurIPS 2022

Guaranteed Discovery of Controllable Latent States with Multi-Step Inverse Models
Alex Lamb*, Riashat Islam*, Yonathan Efroni, Aniket Didolkar, Dipendra Misra, Dylan Foster, Lekan Molu, Rajan Chari, Akshay Krishnamurthy, John Langford.
TMLR 2023

Representation Learning in Deep RL with Discrete Information Bottleneck
Riashat Islam, Hongyu Zang, Manan Tomar, Aniket Didolkar, Md Mofijul Islam, Anirudh Goyal, Samin Yeasar Arnob, Xin Li, Tariq Iqbal, Nicolas Hess, Alex Lamb.

Behavior Prior Representation Learning for Offline Reinforcement Learning
Hongyu Zang, Xin Li, Jie Yu, Chen Lu, Riashat Islam, Remi Tachet Des Combes, Romain Laroche.
ICLR 2023

Offline Policy Optimization with Variance Regularization
Riashat Islam*, Samarth Sinha*, Homanga Bharadhwaj, Samin Yeasar Arnob, Zhuoran Yang, Zhaoran Wang, Animesh Garg, Lihong Li, Doina Precup
Arxiv 2022, The Optimization Foundations for Reinforcement Learning workshop, NeurIPS 2021

Importance of Empirical Sample Complexity Analysis for Offline Reinforcement Learning
Samin Yeasar Arnob, Riashat Islam, Doina Precup
Offline RL Workshop, NeurIPS 2021

Entropy Regularization with Discounted Future State Distributions in Policy Gradient
Riashat Islam, Raihan Seraj, Pierre-Luc Bacon, Doina Precup.
Arxiv 2020, The Optimization Foundations for Reinforcement Learning workshop, NeurIPS 2019

Marginalized State Distribution Entropy Regularization in Policy Optimization
Riashat Islam, Zafarali Ahmed, Doina Precup.
Arxiv 2020, Deep Reinforcement Learning workshop, NeurIPS 2019

InfoBot: Transfer and Exploration via the Information Bottleneck
Anirudh Goyal, Riashat Islam, DJ Strouse, Zafarali Ahmed, Hugo Larochelle, Matthew Botvinick, Sergey Levine, Yoshua Bengio.
ICLR 2019

Provably Efficient Exploration in Policy Optimization with Thompson Sampling
Haque Ishfaq, Zhuoran Yang, Andrei Lupu, Viet Nguyen, Lewis Liu, Riashat Islam, Zhaoran Wang, Doina Precup
Arxiv 2019, ICML RL Theory Workshop 2021

Deep Reinforcement Learning that Matters
Peter Henderson*, Riashat Islam*, Philip Bachman, Joelle Pineau, Doina Precup, David Meger.
AAAI 2018 Science Magazine Blogpost

An Introduction to Deep Reinforcement Learning
Vincent Francois-Lavet, Peter Henderson, Riashat Islam, Marc G. Bellemare, Joelle Pineau
Journal : Foundations and Trends in Machine Learning, 2018

Prioritizing Starting States for Reinforcement Learning
Arash Tavakoli, Vitaly Levdik, Riashat Islam, Petar Kormushev.
Arxiv 2019, Deep RL Workshop NeurIPS 2019

Doubly Robust Off-Policy Actor-Critic Algorithms for Reinforcement Learning
Riashat Islam, Raihan Seraj, Samin Yeasar Arnob, Doina Precup.
Arxiv 2018, Safety and Robustness in RL Workshop, NeurIPS 2019

Discrete off-policy policy gradient using continuous relaxations
Andre Cianflone, Zafarali Ahmed, Riashat Islam, Avishek Joey Bose, William L. Hamilton
RLDM 2019

Re-Evaluate : Reproducibility in Evaluating Reinforcement Learning Algorithms
Khimya Khetarpal, Zafarali Ahmed, Andre Cianflone, Riashat Islam, Joelle Pineau
Reproducibility in ML Workshop, ICML 2018

Transfer Learning by Modelling Distribution over Policies
Disha Shrivastava, Eeshan Gunesh Dhekane, Riashat Islam
Arxiv 2018, Multi-Task and Lifelong Reinforcement Learning, ICML 2019

VFunc : A Deep Generative Model for Functions
Philip Bachman, Riashat Islam, Alessandro Sordoni, Zafarali Ahmed.
Arxiv 2018, ICML Workshop on Prediction and Generative Modelling in RL

Variational State Encoding as Intrinsic Motivation in Reinforcement Learning
Martin Klissarov*, Riashat Islam*, Khimya Khetarpal, Doina Precup
Arxiv 2018, Task Agnostic RL Workshop, ICLR 2019

Off-Policy Policy Gradient Algorithms by Constraining the State Distribution Shift
Riashat Islam*, Deepak Sharma, Komal Teru, Joelle Pineau.
Arxiv 2018

Bayesian Hypernetworks
David Krueger*, Chin-Wei Huang*, Riashat Islam*, Ryan Turner, Aaron Courville.
Arxiv 2018

Deep Bayesian Active Learning with Image Data
Yarin Gal*, Riashat Islam*, Zoubin Ghahramani.
ICML 2017

Research Interests
  • Deep Reinforcement Learning
  • Latent State Discovery (State Abstraction and Representation Learning) and Exploration
  • Offline Reinforcement Learning
  • Optimization Methods for RL
  • Probabilistic Models and Bayesian Deep Learning
  • Deep Generative Models, Approximate and Bayesian Inference
  • Active Learning