|
Reza Asad I'm a PhD candidate at Simon Fraser University, co-advised by Sharan Vaswani and Manolis Savva. Previously, I worked as a data scientist in the San Francisco Bay Area for three years. Before that, I was a research and teaching assistant at the University of British Columbia in Applied Mathematics. I hold a Bachelor's degree in Pure Mathematics from the University of Toronto. |
|
ResearchI'm interested in optimization in ML and RL. My recent work focuses on designing and analyzing RL objectives based on mirror descent principles in the on-policy setting, as well as studying their off-policy variants with entropy regularization. More recently, I have been exploring applications of RL to training LLMs and am actively learning more about this area. |
Papers
Revisiting Actor-Critic Methods in Discrete Action Off-Policy Reinforcement Learning
Optimistic Actor-Critic with Parametric Policies: Unifying Sample Efficiency and Practicality
Fast Convergence of Softmax Policy Mirror Ascent
Surrogate Minimization: An Optimization Algorithm for Training Large Neural Networks with Model Parallelism
3DSSR: 3D Subscene Retrieval
Steiner symmetrization along a certain equidistributed sequence of directions
Cloudmaskgan: A content-aware unpaired image-to-image translation algorithm for remote sensing imagery
Embedded eigenvalues and the nonlinear Schrödinger equation
|