TSVP Talk: "Models of the ventral visual pathway, from neurobiology to artificial networks" by Wyeth Bair

Title: Models of the ventral visual pathway, from neurobiology to artificial networks

Speaker: Wyeth Bair, University of Washington
 
Abstract:
Neuroscientists have studied the physiological properties of single neurons in the brains of highly visual organisms for many decades with the ultimate goal of understanding the circuits and computations that underlie our ability to effortlessly interpret complex scenes and rapidly recognize objects in challenging visual conditions. Much progress has been made in outlining the general principles of organization in the retina and at the first stages of the visual cortex, but it has proven far more challenging to understand, and to model, the middle and higher levels of cortical visual processing. Yet, the past decade has seen a revolution in artificial visual systems arising from machine learning, and now deep artificial neural networks (ANNs) show human-level performance on many vision tasks. This offers a great challenge and a great opportunity for neuroscientists to take advantage of these ANNs to study visual representation. I will compare models created by neuroscientists to those arising from machine learning, and will demonstrate how ANNs can be leveraged by neurophysiologists to better understand the primate visual cortex. At the same time, insights from neurophysiology may be important to guide the next generation of artificial vision systems.

Profile: 
I received my bachelor's degree in Computer Science at Penn State University and my doctorate in Computation and Neural Systems from Caltech, where my PhD work focused on computer modeling of the visual cortex, neural coding and analog VLSI circuits.  As a post-doc at New York University, I studied temporal dynamics in cortical motion and form processing via multi-electrode recordings in non-human primates (NHPs). I started my laboratory at the University of Oxford, UK, funded by the Wellcome Trust and supported by St. John's College, where I studied neural coding in the visual cortex and began to build a general modeling framework for the motion processing pathway from retina to cortical area MT. In 2011, I moved to the University of Washington in Seattle, WA, USA, where I now have a 2- and 3-photon imaging rig for NHPs at the Washington National Primate Research Center (WaNPRC). I have expanded my areas of interest to include mid-level form processing and "Artiphysiology" - the study of artificial neural networks as model visual systems.  In collaboration with the Allen Institute for Brain Sciences, my lab is advancing functional connectomic studies of the primate visual cortex.

Language: English

Target audience: General audience/everyone at OIST and beyond.
Freely accessible to all OIST members and guests without registration.

This talk will also be broadcast online via Zoom:
Zoom link
Meeting ID: 927 0267 3917
Password: 347918

※ Please note that this event may be recorded and the videos uploaded. In addition, photos may be taken during the event. These are intended for publication online (the OIST website, social media, etc.)※