I am a Research Scientist at Google DeepMind in San Francisco, and also an associate professor (status-only) in the Department of Electrical & Computer Engineering at the University of Toronto. I have a broad set of interests within machine learning, but recently I am working on
- Deep Generative Models (in particular Diffusion Models)
- Alignment / Inference / Reasoning in Large Language Models
- Variational Inference and Monte Carlo Methods
- Optimization Methods for Neural Networks
- Computational Optimal Transport
- Intersection of Machine Learning with Information Theory
Previosuly, I was a faculty member at the Vector Institute for Artificial Intelligence, and a Canada CIFAR AI Chair from 2018 to 2024. I completed my PhD (thesis) at the University of Toronto with Brendan Frey in 2018, where I was part of the Machine Learning Group. During my PhD, I interned for Google DeepMind in 2016, and Google Brain in 2015. I completed my Masters (thesis) at the University of Toronto in 2012, and received my Bachelor’s degree from Amirkabir University of Technology, Iran, in 2010.
Email: makhzani_at_cs_toronto_edu
Selected Publications
- Probabilistic inference in language models via twisted sequential Monte Carlo
Stephen Zhao*, Rob Brekelmans*, Alireza Makhzani**, Roger Grosse**
ICML, 2024, (Best Paper Award) - A computational framework for solving Wasserstein Lagrangian flows
Kirill Neklyudov*, Rob Brekelmans*, Alexander Tong, Lazar Atanackovic, Qiang Liu, Alireza Makhzani
ICML, 2024, Also presented in NeurIPS Workshop on Optimal Transport and Machine Learning, 2023 - Action matching: learning stochastic dynamics from samples
Kirill Neklyudov, Rob Brekelmans, Daniel Severo, Alireza Makhzani
ICML, 2023 - Improving mutual information estimation with annealed and energy-based bounds
Rob Brekelmans*, Sicong Huang*, Marzyeh Ghassemi, Greg Ver, Roger Grosse, Alireza Makhzani
ICLR, 2022 - Improving lossless compression rates via Monte Carlo bits-back coding
Yangjun Ruan*, Karen Ullrich*, Daniel Severo*, James Townsend, Ashish Khisti, Arnaud Doucet, Alireza Makhzani, Chris Maddison
ICML, 2021, (Long Oral Talk) - Evaluating lossy compression rates of deep generative models
Sicong Huang*, Alireza Makhzani*, Yanshuai Cao, Roger Grosse
ICML, 2020, Also presented in NeurIPS Workshop on Bayesian Deep Learning, 2019, (Contributed Talk) - Adversarial autoencoders
Alireza Makhzani, Jonathon Shlens, Navdeep Jaitly, Ian Goodfellow, Brendan Frey
ICLR Workshop, 2016