Guy Shani Machine Learning and Applied Statistics guyshani@microsoft.com

I am a researcher at Microsoft Research since August 2007. I graduated at 2007 from Ben Gurion University, Israel, under the supervision of Prof. Ronen Brafman and Prof. Solomon E. Shimony. My dissertation has focused on learning and solving Partially Observable Markov Decision Processes (POMDPs), including both efforts in Reinforcement Learning under partial observability, and solving POMDPs using Point-Based methods. My M.Sc. dissertation, also under the supervision of Prof. Brafman and the guidance of David Heckerman, concerned the application of a Markov Decision Process to a Recommender System.

I am interested in several research areas:

· My main research field is control under uncertainty. I am mainly interested in POMDPs, but also in any other technique that can be used to provide optimal control when actions have stochastic effects and the world features are not fully available. I am interested both in a model-based reinforcement learning, where a model of the environment is learned through trial and error, or through previous recorded interactions with the environment, and in obtaining approximate solutions to such models. My current goal is to find good applications of POMDP techniques to real world problems. Such success stories can help the POMDP community to make a considerable impact on the research community.

· I am also active in the recommender systems research. Recommender systems attempt to suggest items to users, usually customers of an e-commerce web site, trying to enrich the user experience but also to provide value to the site owner. I am interested in new methods to create predictive models of user behavior, in the mining of relevant data for recommendations, in advanced methods for measuring the quality of recommendations, and in any other topic concerning recommender systems.

· I am interested in applications of Machine Learning techniques to real world problems. Specifically, I am mostly interested in problems that arise in the software industry, rather than in the physical world. I believe that there is much potential to apply ML techniques for understanding, managing, and controlling software processes.

Publications:

1) Mining Recommendations From The Web, Guy Shani and Christopher Meek and Max Chickering, The 2^nd International Recommender Systems Conference (RecSys), 2008.

2) Efficient ADD Operations for Point-Based Algorithms, Guy Shani and Pascal Poupart and Ronen Brafman and Solomon E. Shimony, The International Conference on Automated Planning and Scheduling (ICAPS), 2008.

3) Prioritizing Point-Based Solvers, Guy Shani and Ronen Brafman and Solomon E. Shimony, IEEE Transactions on Systems, Man, and Cybernetics, Part B (SMC-B), to appear, 2008.

4) Learning and Solving Partially Observable Markov Decision Processes, Guy Shani, PhD dissertation, Ben Gurion University, 2008.

5) Scaling Up: Solving POMDPs through Value Based Clustering, Yan Virin and Guy Shani and Solomon E. Shimony and Ronen Brafman, AAAI 2007

6) Forward Search Value Iteration For POMDPs - Guy Shani and Ronen I. Brafman and Solomon E. Shimony , IJCAI 2007.

7) Establishing User Profiles in the MediaScout Recommender System, Guy Shani, Lior Rokach, Amnon Meisles, Lihi Naamani, Nischal M. Piratla, David Ben-Shimon, CIDM 2007.

8) A Stereotypes-Based Hybrid Recommender System for Media Items, Guy Shani, Lior Rokach, Amnon Meisles, Yan Gleyzer, David Ben-Shimon, AAAI Workshop on Recommender Systems, AAAI 2007.

9) Prioritizing Point-Based POMDP Solvers - Guy Shani and Ronen I. Brafman and Solomon E. Shimony , ECML 2006.

10) Adaptation for Changing Stochastic Environments through Online POMDP Policy Learning - Guy Shani and Ronen I. Brafman and Solomon E. Shimony, Workshop on Reinforcement Learning in Non-Stationary Environments , ECML 2005.

11) Model-Based Online Learning of POMDPs, Guy Shani and Ronen I. Brafman and Solomon E. Shimony, ECML 2005.

12) Partial Observability Under Noisy Sensors - From Model-Free to Model-Based, Guy Shani and Ronen I. Brafman and Solomon E. Shimony, ICML RRfRL Workshop, ICML 2005.

13) Resolving Perceptual Aliasing In The Presence Of Noisy Sensors, Guy Shani and Ronen I. Brafman  NIPS, 2004.

14) A Survey of Model-Based and Model-Free Methods for Resolving Perceptual Aliasing, Guy Shani, Technical report 05#02 at Department of Computer Science at the Ben-Gurion University in the Negev, November 2004.

15) An MDP-Based Recommender System (M.Sc. Dissertation), Guy Shani .

16) An MDP-Based Recommender System (Journal Version), Ronen I. Brafman, David Heckerman and Guy Shani,  Journal of Machine Learning, 2005.

17) Recommendation as a Stochastic Sequential Decision Problem, Ronen I. Brafman, David Heckerman, and Guy Shani  ICAPS, 2003.

18) An MDP-Based Recommender System, Guy Shani, Ronen I. Brafman, and David Heckerman. UAI, 2002.