Neuroballad: A Neural Circuit Simulation Library

Introduction

Neuroballad is a high level API for simulation of biological neural networks on GPU's, built to facilitate development of circuit-level implementations. Written in Python and on top of Neurokernel/Neurodriver, Neuroballad focuses on user experience and attempts to minimize the time required to execute a desired circuit design.

Neuroballad is in its early stages of development at this time and therefore its use in most circumstances is not recommended.

Installation

Neuroballad is built on and requires a working Neurokernel/Neurodriver installation. For simplicity, we will focus on a Conda-based installation procedure here; other installation methods can be found on the Neurokernel page. To start with Neurokernel, first, add the neurokernel channel to your ~/.condarc file:

channels:
- https://conda.binstar.org/neurokernel/channel/ubuntu1404
- defaults

Afterwards, you can install Neurokernel and Neurodriver as follows:

conda create -n NK neurokernel_deps
source activate NK
# Remember to edit the following line depending on your CUDA version
conda install pycuda=2015.1.3=np110py27_cuda75_0
cd ~
git clone https://github.com/neurokernel/neurokernel.git
cd ~/neurokernel
python setup.py develop
cd ~
git clone https://github.com/neurokernel/neurodriver
cd ~/neurodriver
python setup.py develop

Finally, install Neuroballad:

cd ~
git clone https://github.com/FlyBrainLab/neuroballad
cd ~/neuroballad
python setup.py develop

Getting Started

It is easy to start working with Neuroballad; the following example utilizes three Hodgkin-Huxley neurons connected circularly via alpha synapses:

# Import Neuroballad
from neuroballad import *
# Create a circuit
C = Circuit()
# Create three neurons
C.add([0, 2, 4], HodgkinHuxley())
# Create three synapses
C.add([1, 3, 5], AlphaSynapse())
# Join the nodes together, using an edge list notation
# (i.e. [0,1] implies a directed edge from node 0 to node 1)
C.join([[0,1],[1,2],[2,3],[3,4],[4,5],[5,0]])
# Create a current input for node 0 from t=0.20s to t=0.40s with an amplitude of 40mA.
C_in_a = InIStep(0, 40., 0.20, 0.40)
# Create inputs, similarly, for nodes 2 and 4.
C_in_b = InIStep(2, 40., 0.40, 0.60)
C_in_c = InIStep(4, 40., 0.60, 0.80)
# Use the three inputs and simulate the circuit for a second, in 1e-4s time steps.
C.sim(1., 1e-4, [C_in_a, C_in_b, C_in_c])

Neuroballad includes a number of additional methods for easily dealing with larger systems, the details of which you can find in the documentation.