Abstract: In this talk, we study a model of perceptual integration based on the activities of a sensory neural population. We propose a statistical analysis which allows to recover the scales of this perceptual integration (integration time window, and number of neurons involved) on the basis of experimental samples of neural activity in the population. Formally, this method relies on the quantity of (Fisher) information embedded in random sub-ensembles of the population, depending on their size.
Biography: Adrien Wohrer is a researcher in computational neuroscience, which is the study and formalization of the computations at work in the brain. Most of his research so far has been dedicated to studying the neural basis of perception, both in the visual and tactile systems. After undergrad studies in applied maths and physics, he switched to neuroscience during he PhD, where he designed a model of signal processing in mammalian retinas. He is now a post-doctoral fellow at the Group for Neural Theory of the École Normale Supérieure, in Paris, where he studies sensory encoding in neural populations. The main goal of his work is to recover the typical size of the "functional population" of neurons that vehiculates the percept experienced during a sensory stimulation, and the typical time scale with which this percept is formed
The presentation will be broadcasted on XiMinds.