Neuromorphic hierarchical modular reservoirs

This repository contains code written in support of the work presented in Neuromorphic hierarchical modular reservoirs, as well as data necessary to reproduce the results.

We introduce a simple blockmodeling framework for generating and comparing multi-level hierarchical modular networks and implement them as recurrent neural network reservoirs to evaluate their computational capacity.

Running the simulations

1. Git clone this repository.

2. Git clone my fork of the conn2res toolbox and follow the installation instructions.

3. Running the simulations and plotting the results of the main performance and dynamics analyses additionaly requires the installation of the following dependencies:

   • joblib==1.2.0
   • statsmodels==0.14.0
   • networkx==3.1
   • pingouin==0.5.5

4. To run the simulations, in the command line, type:

python run.py

and pass the relevant flags.

5. To plot the results, simply type:

python plotting.py

Additional analyses are also available in dedicated Jupyter notebooks. All analyses were performed using Python 3.9.0 on a machine running Ubuntu 20.04.6 LTS.

Data

The data folder contains the empirical data used to perform the connectome-informed reservoir computing analyses:

• SC_wei_HCP_s400.npy contains the group-consensus weighted structural connectivity network derived from the HCP dataset (Van Essen et al., 2013; Park et al., 2021), with the 400 cortical nodes defined based on the Schaefer atlas (Schaefer et al, 2018).

• s400_coords.npy contains the coordinates of the nodes in the Schaefer400 atlas (Schaefer et al, 2018).

• parc_timescales.npy contains the map of MEG intrinsic timescales (Shafiei et al., 2023), parcellated according to the Schaefer400 atlas (Schaefer et al, 2018).