Neuromod RNN

This repository contains a JAX implementation of neuromodRNN, which combines e-prop[1] update with diffusion of neurotransmitters (which carry learning signals). The current repostory is still under development and no tutorial notebooks are available.

The jax e-prop implementation of this repository is inspired in the implementation of original e-prop paper authors: https://github.com/IGITUGraz/eligibility_propagation

[1] A solution to the learning dilemma for recurrent networks of spiking neurons G Bellec*, F Scherr*, A Subramoney, E Hajek, Darjan Salaj, R Legenstein, W Maass.

Running

The various implemented methods can be run for different tasks through main.py by setting appropriate flags to override the default configurations. More details will be added in the future.

Installation

The detailed list of required packages is provided in the environment.yml file. You can reproduce the environment using Conda with the following command:

conda env create -f environment.yml -n env_name

Note: The JAX version listed in environment.yml is CPU-only. To run on a GPU, it is necessary to upgrade to the GPU version (see: JAX installation]. To run main.py file within an enivronment with JAX for GPU installed on a node with only CPU available it might be necessary to set the global variable JAX_PLATFORMS=cpu.