BACTAC - Stetten, Strohmann and Howe

Attention is a ubiquitous part of everyday human visual perception, crucially interacting with object recognition. I'll discuss a simple, heuristic, computational model that attempts to capture some key object-based aspects of human visual attention.

The talk will be centered around the "Minimax Probability Machine (MPM)" and the work my advisor Greg Grudic and I were doing on MPM Regression. I will also give some general comments about what being a Research Assistant in Machine Learning is like. For instance, I'll talk a little bit about my experience during the NIPS conference in Vancouver last December and about the relation of my research to graduate school course work.

It has been noted that certain learning tasks can benefit from replacing some study time with equivalent testing time. I will present some of our effects to model this phenomenon using various connectionist frameworks. In addition, I will use these modeling efforts to motivate a discussion of the difficulties and successes of doing computational models of psychological phenomena.