Science Inspires Series Concludes With Pavloski Talk

Posted on 12/3/2014 10:21:38 AM

Raymond Pavloski will close this semester’s Science Inspires lecture series with a presentation titled “The Natural Science of Visual Objects.” The talk will take place on December 4 at 3:30 p.m. in 247 Johnson Hall.

Pavloski will examine visual objects as neural constructs. He has explored this concept for 12 years in his research, and will use his experience to shed light on the relationship between neural activity and visual experience.

This talk will mark the end of this semester’s very successful Science Inspires lecture series. The well-attended series previously featured John Taylor and distinguished guest lecturer Karen Matthews of the University of Pittsburgh.

The abstract for Pavloski’s talk is included below:

The Natural Science of Visual Objects

Although the correspondence between specific aspects of conscious experience and neural activity is the most striking aspect of brain function, the physical basis of this relationship remains a mystery. Researchers typically take a very broad approach to this problem, seeking a small set of neural processes that are common to all types of conscious experience. Although some researchers are very optimistic, this strategy has not led to an accepted theory.

This presentation will describe a different approach that is focused on finding aspects of visual experience and aspects of neural processes that can both be specified using a common formal description. Computer simulations of richly interconnected neural networks are used to emulate known characteristics of the human visual system, and simulation data are compared to vision data.

Results strongly suggest that aspects of recurrent network feedback underlie certain aspects of vision. Early work showed that low-dimensional patterns of feedback emerge in response to retinal input, and that these patterns are characterized by the stability and subjectivity of visual percepts. Current work focuses on the tolerance of a neural network to sufficiently small differences in recurrent feedback generated by different neurons in the network. Simulations reveal that tolerance yields patterns that are characterized by the continuity and unity possessed by visual objects, and suggest that they may also possess additional properties of human vision. Novel predictions and directions for future work will be described.

College of Natural Sciences and Mathematics