I am a PhD candidate at Mila and McGill University, where I am co-advised by Prof. David Meger and Prof. Joelle Pineau. I was previously part of the Autonomous Vehicles Lab at the Department of Mechanical and Mechatronics Engineering at the Univeristy of Waterloo supervised by Prof. Steven Waslander.
I'm interested in designing learning algorithms for robotics applications. My work over the past year has focused on reinforcement learning, especially related to transfer learning in the context of Sim2Real and imitation learning.
Contact : melissafm24 at gmail dot com
PhD in Computer Science, 2020-Present
MASc in Mechanical and Mechatronics Engineering, 2018
University of Waterloo
BSc in Computer Science, 2014
University of St Andrews
Intervention Design for Effective Sim2Real Transfer
Melissa Mozifian, Amy Zhang, Joelle Pineau, and David Meger
Shaping Rewards for Combined Reinforcement and Imitation Learning using Generative Models
Yuchen Wu, Melissa Mozifian, Florian Shkurti
International Conference on Computer Vision on Robotics and Automation (ICRA), 2021
Learning Domain Randomization Distributions for Training Robust Locomotion Policies
Melissa Mozifian*, Juan Camilo Gamboa Higuera*, David Meger, Gregory Dudek
International Conference on Intelligent Robots and Systems (IROS), 2020
Multi-Task and Lifelong Reinforcement Learning Workshop (ICML), 2019
Joint 3D Proposal Generation and Object Detection from
Jason Ku, Melissa Mozifian, Ali Harakeh, Jungwook Lee, Steven Waslander
International Conference on Intelligent Robots and Systems (IROS), 2018
Deep Learning a Quadrotor Dynamic Model for Multi-Step Prediction
Nima Mohajerin, Melissa Mozifian and Steven Waslander
International Conference on Computer Vision on Robotics and Automation (ICRA), 2018
Using deep recurrent neural network architecture to model quadrotor dynamics with autoencoders as a pre-training technique.
Electrical brain activity (EEG) based seizure forecasting systems using machine learning methods. See the challenge
The goal of this project was to develop a system with the ability to detect emotions such as happiness, sadness, surprise and excitement by analysing the facial cues using SVMs. See video
This application uses OpenCV to detect the user's face to isolate the forehead region. Data is collected from this region over time to estimate the user's heart rate by measuring average optical intensity in the forehead location. Physiological data can be estimated this way due to the optical absorption characteristics of (oxy-) haemoglobin See Plethysmographic imaging
I contributed to Idris, a general-purpose functional programming language with dependent types. I implemented a Secure SQL Query system Using Dependent Types.