Simulating recurrent spiking neural networks with cuda.
Clone this repository onto a cuda-capable machine:
git clone https://github.com/Ebanflo42/cuda-rsnn
Compile the shared object file:
nvcc src/simulation.cu -o src/simulation.so -shared -Xcompiler -fPIC -std=c++11
Running this script will simulate a LIF network of 250 neurons for 20000 timesteps with random (uniform) weights and input currents, and then print a subset of the voltages and spikes:
python pytest.py
Note that the voltages are in a very unrealistic range, since no initialization or regularization tricks have been used.