How Neural Network Structure Shapes Dynamics: In Theory & In Vivo
- Nicole Sanderson (Pennsylvania State University, USA)
Abstract
(Part I) Neuromodulation is a mechanism that can alter the dynamics of a neural network without changing the underlying network’s connectivity structure. This can be modeled with simple recurrent neural networks. Using threshold-linear networks (TLNs), we investigate the extent to which neuromodulation alone can change a network’s dynamics as well as how a network’s connectivity constrains the possible dynamics achievable via neuromodulation. We show that the bifurcation diagram over the neuromodulatory phase space can be described via system of overlapping simplicial cones.
(Part II) Using single-photon calcium imaging, we record for an hour the activity of ~1000s of neurons in zebrafish larvae optic tectum in the absence of stimulation. We observe spontaneous activity of neuronal assemblies that are both functionally coordinated and localized in the optic tectum. We analyze the neural correlations using topological signatures that identify distinct correlation structures during spontaneous activity.