Contextual effects in cortical visual processing and deep neural networks

Neural responses and perception of visual inputs strongly depend on the spatial context, i.e., what surrounds a given object or feature. I will discuss our work on developing a visual cortical model based on the appealing hypothesis that neurons represent inputs in a coordinate system that is matched to the statistical structure of images in the natural environment. The model generalizes a nonlinear computation known as normalization, that is ubiquitous in neural processing. In our framework, normalization is flexible and recruited only when an image is inferred to contain dependencies and muted otherwise. The model can capture some spatial context effects in cortical neurons. I will further discuss how we are incorporating such nonlinearities into deep neural networks.

Odelia Schwartz is an Associate Professor of Computer Science at the University of Miami. Her research is at the intersection of machine learning and the brain sciences. The main focus has been building computational models of how the brain makes sense of visual scenes, resulting in perception and cognition. She received her Ph.D. in the Laboratory for Computational Vision at New York University and an MS in the Computational NeuroEngineering laboratory at the University of Florida, and did her postdoctoral work in the Computational Neurobiology Laboratory at the Salk Institute. Her research has been funded by the NSF, the NIH, the Army Research Office, a Google Faculty Research Award, and an Alfred P. Sloan Research Fellowship Award.

https://yorku.zoom.us/j/96807031417?pwd=RjhxaEFZaWJlWUVxSzVYa2o4VHVWQT09