Emtiyaz Khan

Emtiyaz Khan is a Researcher at the RIKEN Center for AI Project in Tokyo, Japan. His work focuses on Bayesian deep learning, optimization, and variational inference methods. He leads research on the Bayesian Learning Rule framework, which bridges deep learning optimization with Bayesian principles.

His research interests include developing scalable Bayesian methods for large neural networks, uncertainty quantification in deep learning, optimization algorithms (natural gradients, variational inference), and applications to foundation models. Key areas are efficient adaptation methods, model sensitivity analysis, and Bayesian principles for deep learning.

Khan's publications demonstrate strong focus on Bayesian deep learning, optimization techniques, and uncertainty estimation, with applications ranging from large-scale models (GPT-2, ImageNet) to theoretical foundations of variational inference.

He leads the Team Approx-Bayes research group focused on approximate Bayesian inference methods and maintains collaborations through JST CREST-ANR and Kakenhi grants.