Mohammad Gheshlaghi Azar
Rehabilitation Institute of Chicago
345 E Superior Street
Attn: Kording Lab , Rm 1479
Chicago, IL 60611
Email: mohammad.azar (AT) northwestern [DOT] edu
Phone: 312-238-6515 Google Scholar Profile
I am a postdoc with the Rehabilitation Institute of Chicago, Northwestern University (NU). My main area of research is machine learning and statistics. In particular I am interested in applying the-state-of-the-art ML algorithms to solve challeging neuroscience problems such as adaptive and unsupervised bain machine interface (BMI).
My main research interest is to develop efficient machine learning methods which can scale up to large-scale, non-stationary domains.
Particular interests include:
Large-scale non-convex optimization, optimistic optimization with bandit feedback, X-armed bandits.
Transfer of knowledge and domain adaptation in multi-armed bandits and reinforcement learning.
- Finite-time analysis of learning algorithms (Empirical risk bounds, regret bounds, PAC sample-complexity bounds).
- Spectral methods to estimate latent variable models (topic models, HMMs, POMDPs).
- Trade-off between exploration and exploitation (PAC-MDPs and upper-confidence bound methods).
- Application of Machine Learning and reinforcement learning to adaptive brain-machine Interface (BMI).