Brian 2 neural network simulations written as an educational project during the OIST Computational Neuroscience Course in Okinawa, Japan. The networks aim to reproduce and extend results from
Rosenbaum, Robert, et al. "The spatial structure of correlated neuronal variability." Nature neuroscience 20.1 (2017): 107.
To run the network simulation a Python 3 environment with Brian 2 is required. A recommended way of installing the dependencies is for example using virtualenv and pip:
#!/bin/sh
virtualenv -p python3 /home/$USER/env/ocnc-net
source /home/$USER/env/ocnc-net/bin/activate
pip install brian2
pip install matplotlibIn blncd_net a balanced network driven by external Ornstein-Uhlenbeck noisy currents is simulated. This is an (incomplete) attempt to reproduce Figure 1 of Rosenbaum et al. (2017).
Run the simulation via
source run.sh
In blncd_net/img/ the following output is expected
In ffwd_rec a recurrent network with neuron distributed on a grid are driven by spatially distributed feedforward input similar to Figure 2-4 in Rosenbaum et al. (2017).
Run the simulation via
source run.sh
In ffwd_rec/img/ the following output is expected
In plst_net the model of Rosenbaum et al. (2017) is extended to include spike-timing dependent plasticity (STDP) and homeostatic plasticity.
Run the simulation via
source run.sh
In plst_net/img/ the following output is expected
Lastly, in struct_plast the model is further extended to include structural plasticity in which connection between neuron rewire over time.
The struct_plst.py file gives an idea how structural plasticity (which is not at the moment natively supported) can be implemented in Brian 2.
The network simulation can be run via
source run.sh
Further analysis on this network was not done, however.
Many thanks go to Marcel Stimberg who supervised and gave much helpful input to this project during the summer school.