Representing and reasoning about uncertain knowledge
Event Time : 11:00:00
Event Location : Prof. Rasmy Seminar Hall
Event Oragnizer : Faculty of Computers and Information
Abstract: This talk is primarily concerned with what we can do with uncertain and sometimes incomplete knowledge about a system or phenomenon of interest. The first part of the talk gives an overview of graphical models, a compact, intuitive and widely applicable representation of probabilistic knowledge. We will discuss different kinds of graphical models, demonstrate with simple examples and real life applications, and give a brief idea of how the model can be used to answer queries about the system/phenomenon.
The second part of the talk discusses how we can reason under uncertainty to make rational decisions. We will explore models based on Markov decision processes, applications and solution methods. Time permitting, we will also discuss the challenges we face when we have multiple agents making decisions under uncertainty. This talk does not require any background in probabilistic or decision making models. Biography: Hala Mostafa obtained her BSc and MSc from FCI-CU and PhD from the University of Massachusetts in the field of Artificial Intelligence. Her areas of interest include single and multi-agent decision making under uncertainty, graphical models and distributed constraint satisfaction. She is currently a Research Scientist at BBN Technologies. cs.umass.edu/~hmostafa