This talk focuses on the need for and meaning of PRAM Algorithms for computer systems that learn from and react to data from sensors. Possibility theory is an axiomatic theory similar to probability that provides different mechanisms for ambiguity and robustness. Robust algorithm is an often used phrase that has become almost meaningless. There are several definitions given in the literature and sometimes no definition is given. However, statistical robustness is very important for algorithms used to process sensor data in somewhat unconstrained environments. Ambiguous patterns are patterns of measurements that could easily represent different classes of objects. The talk will provide an overview of possibility and robustness theory and relate that theory to the machine learning topic of manifold learning. Discussions of robustness properties of popular classifiers such as Support Vector Machines, Deep Learning Networks, and Bayesian methods will be given. Applications will be described in the fields including Hyperspectral Image Analysis.
Dr. Gader is a Professor and former chair of the Computer &Information Science & Engineering (CISE) department and is also affiliated with Environmental Engineering at the University of Florida. He has been a Senior Research Scientist at Honeywell; Research Engineer & Manager at the Environmental Research Institute of Michigan. He has published over 100 refereed journal papers and is a Fellow of the IEEE.