BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Department of Psychology - ECPv4.9.3.2//NONSGML v1.0//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-WR-CALNAME:Department of Psychology
X-ORIGINAL-URL:http://psych.wisc.edu
X-WR-CALDESC:Events for Department of Psychology
BEGIN:VEVENT
DTSTART;TZID=UTC+0:20190418T160000
DTEND;TZID=UTC+0:20190418T170000
DTSTAMP:20260415T213042
CREATED:20190110T222913Z
LAST-MODIFIED:20190416T140631Z
UID:3253-1555603200-1555606800@psych.wisc.edu
SUMMARY:Colloquium Series -- Frank Tong of Vanderbilt University (April 18th)
DESCRIPTION:“Understanding the neural computations that underlie human vision” \nEach morning\, we open our eyes and re-experience the scientific miracle of vision. In this talk\, I will discuss two of my lab’s major endeavors to investigate the neurocomputational basis of visual perception. First\, I will describe ongoing studies that pit human observers and deep neural networks (DNNs) at challenging near-threshold tasks of object recognition. We find that people are much better at recognizing objects in extreme levels of visual noise than state-of-the-art DNNs. However\, we have successfully developed methods to create noise-robust DNNs that outperform human observers and make errors that share greater resemblance to human performance. Our findings demonstrate that neural architectures with feedforward processing alone can attain robustness to visual noise and clutter\, without recourse to top-down mechanisms of attentional feedback. In the second part of my talk\, I will discuss our fMRI studies of the lateral geniculate nucleus\, and its potential role of figure-ground processing and predictive visual coding. The LGN is classically portrayed as a relay station that simply serves to transfer information from retina to cortex; however\, such an account fails to explain why the LGN receives far more afferents from the visual cortex than from the retina. We find fMRI evidence of orientation processing in the human LGN as well as attentional modulation of orientation responses. Moreover\, this subcortical region shows enhanced responses to perceptual figures that deviate from the background. Further experiments suggest that figure-ground modulation in the LGN is attributable to automatic top-down feedback from binocular visual cortex. The LGN’s sensitivity to deviations between figure and surround provides intriguing new evidence to support neurocomputational models of predictive coding. I will conclude by discussing how neuroscientific experiments\, coupled with advanced modeling using DNNs\, can help shed light on the functional role of feedforward and feedback processing in the human visual system. \nTalk held in room 121 of Brogden Psychology Building. \nRefreshments will be provided at 3:50 pm. \nFor information regarding the talk contact uwcsc@googlegroups.com. \n
URL:http://psych.wisc.edu/event/colloquium-series-frank-tong-april-18th/
LOCATION:TBA
CATEGORIES:Biology of Brain and Behavior,Cognitive and Cognitive Neuroscience,Colloquium Series,Departmental,Graduate,Undergraduate
ATTACH;FMTTYPE=image/jpeg:http://psych.wisc.edu/wp-content/uploads/2019/01/frank_tong_400.jpg
END:VEVENT
END:VCALENDAR