Keynotes | K&Ts | GACs | Talks | Posters | Search
Poster A in Poster Session A: Tuesday, August 4, 9:30 – 11:15 am, Kimmel Center, Shorin & Rosenthal Rooms
A tale of two tails: Preferred and anti-preferred natural stimuli in visual cortex
Rabia Gondur1, Patricia L. Stan2, Matthew A. Smith2, Benjamin R. Cowley1; 1Cold Spring Harbor Laboratory, 2Carnegie Mellon University
Presenter: Rabia Gondur
A fundamental quest in neuroscience is to find the preferred stimulus of a sensory neuron. The prevailing notion is that a visual neuron primarily responds to a single preferred visual feature, like an oriented edge or object identity, resulting in a “one-tailed” distribution of responses to natural images. However, we instead find “two-tailed” response distributions of macaque visual cortical neurons, suggesting that real neurons have both preferred and anti-preferred stimuli. We verified the existence of anti-preferred stimuli by recording responses from macaque V4 neurons to model-optimized stimuli. We find that these anti-preferred stimuli are essential for describing a neuron’s tuning, as responses to both preferred and anti-preferred images are needed to predict a neuron’s response to natural images. Moreover, humans performing a psychophysics task relied on anti-preferred images to infer a V4 neuron’s stimulus tuning; this was not the case for inferring the tuning of internal units from a deep neural network. The features of preferred and anti-preferred images were seemingly unrelated, suggesting that V4 neurons encode a broader range of features beyond those they “prefer”, enriching the V4 population’s representational basis. To encourage further experiments investigating anti-preferred images, we developed a tool called ImageBeagle that efficiently “hunts” through millions of natural images, for real-time experiments. Overall, our work embarks on a new quest in neuroscience to search for anti-preferred stimuli along the visual stream, as well as to better understand how feature selectivity arises in the visual cortex and deep neural networks.
Topic Area: Computational Models of Vision & Visual Cortex