peter-adepoju/sbi_for_stochastic_hh_models

My Master's thesis at AIMS South Africa # Efficient Inference of Stochastic Models of Single Neurons with Multifidelity Simulation-Based Inference This repository contains the official source code and results for the Master's thesis titled "Efficient Inference of Stochastic Models of Single Neurons with Multifidelity Simulation-Based Inference" by Peter Oluwafemi Adepoju, submitted in partial fulfillment of a structured master's degree at the African Institute for Mathematical Sciences (AIMS), South Africa, a degree jointly awarded by the University of Cape Town. This work implements and validates a Multifidelity Neural Posterior Estimation (MF-NPE) pipeline for efficiently inferring the biophysical parameters of stochastic Hodgkin-Huxley neuron models. --- ## Abstract The voltage across the membranes of individual neurons is inherently stochastic, primarily due to the random transitions between conformational states of membrane ion channels. Accurately modelling these phenomena requires computationally intensive simulations, making parameter inference impractical for many neuroscience applications. This project addresses this challenge by developing a multifidelity simulation-based inference (SBI) approach. We combine inexpensive, low-fidelity simulations with a limited budget of costly, high-fidelity simulations to train a neural density estimator. Our results show that this approach significantly outperforms traditional methods, achieving higher accuracy while requiring orders of magnitude fewer high-fidelity simulations. The study presents a validated, open-source framework that enhances parameter inference in computational neuroscience. --- ## Setup and Installation ### 1. Prerequisites - Python 3.10 or higher - Git ### 2. Clone the Repository Clone this repository to your local machine: ```bash git clone github.com cd sbi_for_stochastic_hh_models ``` ### 3. Install Dependencies It is highly recommended to use a virtual environment to m …