Mehrab Hamidi

I’m passionate about understanding how neural networks and interpretable models make their decisions. My research primarily focuses on deep learning theory, mechanistic interpretability, and explainability.

I am currently a PhD student at the Université de Montréal and a researcher at Mila, working under the mentorship of Professors Ioannis Mitliagkas and Aristide Baratin. My academic journey includes a master’s degree in computer science from McGill University under the supervision of David Rolnick and an undergraduate degree in computer science with a minor in mathematics from Sharif University of Technology.

My journey began with a fascination for the intersection of computer science and biology. I was captivated by how algorithms could solve extraordinarily complex biological problems, such as those tackled in the Human Genome Project. This undeniable connection led me to machine learning and, eventually, deep learning. Over time, I realized that while these models hold immense potential, they often fall short in critical applications, particularly in the medical field. For them to be truly impactful, they need to be understandable and explainable. This realization sparked my interest in deep learning theory, where I started exploring the representations and decision-making processes of neural networks from statistical, geometrical, and topological perspectives. Recognizing the complexity of this challenge, I delved into mechanistic interpretability questions to gain deeper insights into the internal structure of neural networks.

Beyond research, I thrive on collaboration and helping my community. I’ve taken on teaching, mentoring, and leadership roles that allow me to share knowledge and inspire others. When I’m not immersed in code or papers, you’ll find me hiking, playing tennis, exploring new music, or enjoying time at the gym. I also play the violin, kamancheh, and saxophone.

If you’re interested in discussing ideas or collaborating on exciting projects, don’t hesitate to reach out! :)