Work

Neural Posterior Estimation of hierarchical models in neuroscience

Year: 2022
Type: conference-paper
Abstract: Understanding the bio-physical mechanisms underlying complex neuronal phenomena is the main focus of experimental researchin neuroscience. When trying to computationally replicate realistic brain signals, it typically involves inferring the parameters of stochastic nonlinear models. This task becomes particularly challenging when the coupling of certain parameters leads to intrinsically indeterminate models.We present a method that is capable of removing such indeterminacy by extending recent developments in simulation-based inference (SBI) tohierarchical Bayesian models. The idea is to exploit additional information conveyed by an auxiliary set of observations sharing the same globalparameters, in order to provide more accurate estimates of the coupled parameters. We demonstrate this method on the well known Jansen & RitNeural Mass Model and use it to investigate what extra information best improves the parameter estimates on EEG signals
Source: HAL
Language: English
Open Access status: green
OpenAlex ID: W4309371531
FWCI: 0.963
Cites:
Cited by: 3
Related to: 10