Event tree analysis of GLIF networks
Coarse-grained dynamics and input discriminability in spiking networks
Coarse-Grained Event Tree Analysis of GLIF Neuronal Networks Department of Mathematics, University of South Carolina. Advisor: Dr. Yi Sun, May 2024 – May 2025.
Generalized leaky integrate-and-fire (GLIF) networks sit between the tractability of LIF models and the biophysical detail of Hodgkin–Huxley dynamics. This work asks how much of a network’s computational behavior can be recovered from the coarse structure of its spiking events alone.
Event tree framework
- Built a Python event tree analysis (ETA) framework for GLIF networks to quantify network dynamics and input discriminability.
- Used conditional probability distributions over event chains to measure sensitivity to small input perturbations.
Result
Synaptic conductance and Gaussian noise were shown to critically shape excitability thresholds, and therefore the network’s capacity to discriminate nearby inputs.
Earlier work
This followed an earlier project (Feb 2023 – May 2024) implementing the GLIF model itself and a simulation pipeline for neuronal responses under variable noise and synaptic coupling, supported by the UREP Grant and a Magellan Fellowship.