Mohammad Ghavamzadeh

• Researcher   INRIA Lille - Nord Europe, Team SequeL, France
  (2008 - Present)

• Postdoctoral Fellow   University of Alberta, Canada  (2005 - 2008)

• Ph.D. in Computer Science   University of Massachusetts Amherst, USA  (2001 - 2005) 

Recent news...

• Three papers got accepted at ICML-2010. "Bayesian Multi-Task Reinforcement Learning" with Alessandro Lazaric, "Analysis of a Classification-based Policy Iteration Algorithm" and "Finite-Sample Analysis of LSTD" with Alessandro Lazaric & Rémi Munos. 

• Our paper (with Shalabh Bhatnagar and Rich Sutton) on "Natural Actor-Critic Algorithms" was published at Automatica. Its longer version is available as a UAlberta Tech-Report [pdf]. In this paper, we present four new actor-critic algorithms (three using natural gradient and one using regular gradient) with convergence proofs.

• I am co-organizing, with Yaakov Engel, Shie Mannor, and Pascal Poupart, a workshop on "Model Uncertainty and Risk in Reinforcement Learning" at  NIPS 2008 on December 13, 2008.

• I am moving to France to join INRIA Lille - Nord Europe, Team SequeL as a researcher in November 2008.
  

• I am co-teaching, with Csaba Szepesvari, "Reinforcement Learning in Artificial Intelligence" in Fall 2007.

• I am co-organizing, with Pascal Poupart and Yaakov Engel, a tutorial on "Bayesian Methods in Reinforcement Learning" at  ICML 2007 on June 20, 2007.