Speakers > Andrea Benucci
Riken center for brain science Andrea Benucci is a Team Leader (equivalent to Assistant Professor) at RIKEN Center for Brain Science in Japan, near Tokyo. Andrea holds a Bachelor’s degree in physics from the University of Padova in Italy, a Master degree in theoretical neuroscience from the International School for Advanced Studies (SISSA) in Trieste, and a PhD also in theoretical neuroscience from the ETH-Zurich in Switzerland. Before opening his laboratory at RIKEN across 2013-2014, he was a postdoctoral researcher at the Smith-Kattlewell Eye Research Institute in San Francisco, working with Matteo Carandini, and then he moved with Carandini to London, at the UCL Institute of Ophthalmology, as a Senior Research Associate. Andrea’s research focuses on the principles that govern the dynamics of large populations of neurons in the visual cortex of mice, and on how the circuit dynamics support the computations that underlie visual perception and perceptual decisions.
Motor-related signals support localization invariance for stable visual perception
Our ability to perceive a stable visual world in the presence of continuous movements of the body, head, and eyes has puzzled researchers in the neuroscience field for a long time. We reformulated this problem in the context of hierarchical convolutional neural networks (CNN) whose architectures have been inspired by the hierarchical signal processing in the mammalian visual system. We examined perceptual stability as an optimization process in networks trained to accurately classify images in the presence of self-generated movements (simulated saccades). Motor-related activations multiplexed with visual inputs along overlapping convolutional layers and carrying information about self-generated movements, aided classification invariance of shifted images by making the classification faster to learn and more robust relative to input noise. Classification invariance was reflected in activity manifolds associated with image categories emerging in late CNN layers and with network units acquiring movement-associated activity modulations as observed experimentally during saccadic eye movements. Our findings provide a computational framework that unifies a multitude of biological observations on perceptual stability under optimality principles for image classification in artificial neural networks.
|
