CellTypeMeanField

This repository contains code and data for the study "Inhibitory cell type heterogeneity in a spatially structured mean-field model of V1" by Kim and Choi (https://doi.org/10.1101/2025.03.13.643046).

Fig. 1e: Rastor plot under Gaussian input

First compile the simulation code in Matlab using mex LIF4pop.c.

Then run the simulation with runSimSpikes('param_base', 300000, 42).

Use Fig1e.ipynb to plot raster.

Fig. 2: Mean-field model validation

To find mean-field model fixed points with network size 3000000, run python param_base 300000.

To obtain firing rates quickly (without saving spikes), use runSimFR('param_base', N, seed).

Follow Fig2.ipynb to compare mean-field model with spiking simulations and analytical solutions. Need to run for multiple sizes before Fig2b-c can be plotted.

Fig. 3: Stability analysis

Run Fig3.ipynb for bifurcation/stability analysis.

Fig. 4: Spatiotemporal dynamics

To run simulations, first create a batch of parameter files using python create_batch.py param_base. Then run runSimSpike`` for all sigma_e values. Run Fig4.ipynb``` for analysis of spatiotemporal features from spiking.

Fig. 5: Gain modulation and linear response theory

To compute the mean-field solutions under stimulation conditions, run

python gain_fp_current.py gain_current control

for control condition (no inhibitory neuron modulation) and

python gain_fp_current.py gain_current pv_pos 0.5 2.0

for modulatory simulations. In the above example, PV is stimulated at position 0.5 with strength factor 2.0. The visual input is given at 0.5, so this simulates proximal stimulation. Use 0.0 for distal. The strength factor is modulates the stimulus to the modulating inhibitory neuron. The strength factor multiplied by the spontaneous input current is the raw stimulation strength.

For conductance-based model, use

python gain_fp_cond.py gain_cond pv_pos 0.5 2.0

and so on.

Use Fig5.ipynb for plots and linear response theory analysis.