TU GRAZ Informatikkolloquium
It is a pleasure to invite you to the following talk of Prof. Jason MacLean from The University of Chicago:
Date:
Monday, 4th of March 2019 at 14.00
Place:
Seminar room IC01074 – Institute of Theoretical Computer Science (IGI) – Graz University of Technology – Inffeldgasse 16b – 1st floor
Title:
Decoding the functional networks of cerebral cortex: A refinement of the Hebbian Assembly
Abstract:
Understanding shared trial-to-trial variability in V1 neuronal dynamics is key to developing a complete description of the encoding processes of visual stimuli in neocortex. Trial-averaged neuronal response profiles indicate that some neurons respond preferentially to specific statistical features in stimuli but it remains unclear whether single trial co-variability between neurons degrades the stability of sensory representation or indicates the existence of a deeper organizational principle. We analyze two-photon calcium imaging data from populations of layer 2/3 excitatory neurons in awake mouse visual cortex during presentations of drifting gratings and summarize the population’s pairwise correlation structure as a sparse functional network. Using a generalized linear model we relate each neuron’s single trial activity to the stimulus, mouse running speed, and use the experimentally measured partial pairwise correlations as coupling coefficients between neurons. We find that single trial prediction of neuronal activity improves when conditioning on the coupling coefficients, i.e. functional group, of each neuron in comparison to predictions from stimulus, or running speed. Specifically, we find that accurate predictions are driven by strong, positively co-varying, synchronously active neurons that collectively form a functional group a modern analog to Hebb’s assembly. Additionally, we show that Bayesian decoding of the stimulus improves with knowledge of the functional group. These results suggest that specific local population interactions in visual cortex are critical for predicting single trial dynamics.
Assoc. Prof. Jason MacLean
The University of Chicago
Department of Neurobiology