Cosyne 2007 Workshops
February 26-27, 2007
The Canyons, Utah
Workshop Title
How can we understand shape coding in higher level visual areas?
Organizer(s)
Anitha Pasupathy (MIT): anitha@MIT.EDU
Ed Connor (Johns Hopkins Univ): connor@jhu.edu
Abstract
Shape coding is extremely difficult to address experimentally and theoretically, because shape information in visual images is so high-dimensional, variable, and implicit. We want to speculate about what novel experimental and analytical methods might finally make this question tractable. The emphasis will be on actually discovering the neural code for shape (as opposed to important but distinct issues like maximum information, invariance, effects of learning or attention). The major experimental challenge is adequately sampling neural responses across the virtually infinite domain of object shape. The major theoretical challenge is quantifying shape information conveyed by neurons and neural populations. Speakers are encouraged to critique the limitations of current approaches and consider radically new directions, possibly based on analogies to other problems or fields.
Speakers
| Ed Connor (Johns Hopkins University) | Adaptive stimulus mutation to explore neural tuning in the 3D shape domain |
| Jack Gallant (UC Berkeley) | Statistical issues in studies of higher vision: inherent problems and (some) solutions |
| Roozbeh Kiani (U Washington) | Representation of Object Category Structure by ResponsePatterns of Neuronal Population in Monkey Inferior Temporal Cortex |
| David Leopold (NIH) | Norm based coding of faces in the monkey inferotermporal cortex |
| Mike Lewicki (CMU) | A theoretical approach to shape representation based on natural scene statistics |
| Fei-Fei Li (Princeton) | Modeling real-world objects: a statistical approach |
| Tony Movshon (NYU) | Can understanding the analysis of motion il luminate the analysis of form? |
