Romina Abachi
I have recently completed my MSc. from Computer Science (Artificial Intelligence) at the University of Toronto and Vector Institute, supervised by Profs. Amir-massoud Farahmand and Sheila McIlraith. Broadly, my research interests span reinforcement learning and practical and safe RL algorithms.
Previously, I did my Master's in Electrical and Computer Engineering at the University of Toronto, where I was advised by Amir-massoud Farahmand and Brendan Frey.
I did my Bachelor's in Electrical Engineering at the University of Toronto, as well. There, I worked in the Energy Systems lab, where I was advised by Prof. Olivier Trescases.
Previously, I have done internships at Borealis AI, where I studied Adversarial attacks and certified robustness methods, Qualcomm , and Toronto Hydro.
Email  / 
CV  / 
Google Scholar  / 
Github
|
|
Research
My research focuses on Model-Based Reinforcement Learning (MBRL) and safe RL algorithms, including risk-sensitive and robust planning. I am interested in algorithms grounded in theory and directed by practical and real-world problems.
|
|
Optimistic Risk-Aware Model-based Reinforcement Learning
Romina Abachi,
Amir-massoud Farahmand
European Workship in Reinforcement Learning, 2022
PDF
|
|
VIPer: Iterative Value-Aware Model Learning on the Value Improvement Path
Romina Abachi*,
Claas Voelcker*,
Animesh Garg,
Amir-massoud Farahmand
(*equal contribution)
Decision Awareness in Reinforcement Learning Workshop, ICML, 2022
|
|
Policy-Aware Model Learning for Policy Gradient Methods
Romina Abachi,
Mohammad Ghavamzadeh,
Amir-massoud Farahmand
arXiv, 2020
arXiv
/
code
/
short version
|
|
Control-Oriented Model-Based Reinforcement Learning with Implicit Differentiation
Evgenii Nikishin,
Romina Abachi,
Rishabh Agarwal,
Pierre-Luc Bacon
AAAI, 2022
arXiv
|
|