anticache

ICS, CNE, HCR Research Seminar on "Bayesian inference for parametric receptive-field models"

17.12.2018


ICS, CNE, HCR Research Seminar on "Bayesian inference for parametric receptive-field models" by Giacomo Bassetto Computational Neuroengineering group (CNE).

19 December 2018, 16:30h, ICS Karlstraße 45, 6. Floor, room 6009

Abstract: Estimating neural receptive fields is an important step towards understanding how the external world is represented by sensory neurons. Receptive fields often have to be estimated from very limited experimental observation times. Here, we present a Bayesian approach for estimating receptive fields in low-data regimes, and for quantifying and visualising how well receptive field properties are constrained by data. Our approach is based on using Markov-Chain Monte Carlo sampling to perform Bayesian inference directly on the parameters of receptive-field models. We exploit the fact that, for many sensory areas, there are canonical parametric models of receptive field shapes which can be used to constrain the space of receptive-fields which can explain the data. While such models will not be able to capture all nuances of receptive fields, they can be useful for obtaining a fast characterization of receptive field properties of interest, e.g. position and preferred orientation. We show how this approach makes it possible to reliably identify sensory receptive fields using a small number of spikes, and to quantify estimation uncertainty by inspecting the posterior distribution. We demonstrate the effectiveness of this approach on a variety of receptive field models on both simulated and neurophysiological data, with a particular focus on simple cell models of neurons from primary visual cortex.

Bio: Giacomo Bassetto is a Ph.D. candidate at the Department of Electrical Engineering at TUM, within the Computational Neuroengineering group (CNE). His research is supervised by Prof. Jakob Macke. His research interests include statistical modelling of neural population dynamics and Bayesian inference. He received his Master in Bioengineering from the university of Padova (Italy).