Time+Place: | Tuesday 04/11/2014 14:30 Room 337-8 Taub Bld. |

Title: | Robust Probabilistic Inference |

Speaker: | Yishay Mansour - Colloquium Lecture http://www.math.tau.ac.il/~mansour/ |

Affiliation: | School of Computer Science, Tel-Aviv University |

Host: | Erez Petrank |

Probabilistic Inference is the task of given a certain set of observations, to deduce the probability of various outcomes. This is a very basic task both in statistics and in machine learning. Robust probabilistic inference is an extension of probabilistic inference, where some of the observations are adversarially corrupted. Examples of where such a model may be relevant are spam detection, where spammers try adversarially to fool the spam detectors, or failure detection and correction, where the failure can be modeled as a ''worse case'' failure. The framework can be also used to model selection between a few alternative models that possibly generate the data. Technically, we model robust probabilistic inference as a zero-sum game between an adversary, who can select a modification rule, and a predictor, who wants to accurately predict the state of nature. Our main result is an efficient near optimal algorithm for the robust probabilistic inference problem. More specifically, given a black-box access to a Bayesian inference in the classic (adversary-free) setting, our near optimal policy runs in polynomial time in the number of observations and the number of possible modification rules. This is a joint work with Aviad Rubinstein and Moshe Tennenholtz Bio: Prof. Yishay Mansour is a Professor of Computer Science at Tel-Aviv University. He received his PhD from MIT in 1990, following it he was a postdoctoral fellow in Harvard and a Research Staff Member in IBM T. J. Watson Research Center. Prof. Mansour joined Microsoft Research in Israel in 2014. Before that he held visiting positions with Microsoft, Bell Labs, AT&T research Labs, IBM Research, and Google Research. He has mentored start-ups, e.g., Riverhead, which was acquired by Cisco, Ghoonet and Verix. Prof. Mansour has published numerous journal and proceeding papers in various areas of computer science with special emphasis on communication networks, machine learning, and algorithmic game theory, and has supervised over a dozen graduate students in those areas. Prof. Mansour is currently an associate editor in a multiple distinguished journals and has been on numerous conference program committees. He was the program chair of COLT (1998) and serves on the COLT steering committee. Deserts will be served from 14:15 Lecture starts at 14:30