Bio

I am a Deep Learning Researcher at Microsoft Research (MSR) in Montreal, an Adjunct Professor at the Universtiy of Montreal, and an associate member of Mila. My research focuses on using mutual information estimation and self-supervision in representation learning for applications in computer vision, natural language processing, and RL. My long-term goal is to use machine learning to learn representations that help with scientific discovery.

I earned my Ph.D. at the University of New Mexico where I focused on Unsupervised Learning with applications to Neuroimaging. I then did a postdoc under Yoshua Bengio at Mila as an IVADO “distinguished researcher” fellow, focusing on adversarial learning.

• To prospective students: I can host interns and co-supervise students at Mila. Due to limited time, however, I can only possibly take students if they are far enough along in their research. Publications (minimum Workshop at a major conference, e.g., NeurIPS, ICML, ICLR) with work related to representation learning is a must. If you are interested, send me your list of relevant publications as well as current research ideas that overlap with my focuses above. I’m less concerned about which University you attended, your rank, or your GPA. Demonstration of coding ability is a plus.

• December 2018: Two papers accepted to ICLR 2019, both related to Deep InfoMax (DIM)! Learning Deep Representations by Mutual information Estimation and Maximization (a.k.a. Deep InfoMax or DIM) was accepted as an oral presentation (top 1.5% of submissions). Deep Graph InfoMax was accepted as a conference poster.
• October 2018: A paper accepted to AAAI on better mode coverage for generative adversarial networks (GANS) called Online Adaptive Curriculum Learning for GANs.
• September 2018: Two workshops papers accepted to NeurIPS 2018. One is on learning representations.