NeuroMLlite: a common framework for reading/writing/generating network specifications based on NeuroML

NeuroMLlite is in active development. This will evolve into a framework for more portable, concise network specifications which will form an important part of NeuroML v3.

For some more background to this package see here: https://github.com/NeuroML/NetworkShorthand.

Examples

The best way to see the currently proposed structure is to look at the examples

Ex. 1: Simple network, 2 populations & projection

JSON | Python script

Can be exported to:

• NeuroML 2 (XML or HDF5 format)
• Graph (see above)

Ex. 2: Simple network, 2 populations, projection & inputs

JSON | Python script | Generated NeuroML2

Can be exported to:

• NeuroML 2 (XML or HDF5 format)
• Graph (see above)

Ex. 3: As above, with simulation specification

JSON for network | JSON for simulation | Python script | Generated NeuroML2 | Generated LEMS

Can be exported to:

• NeuroML 2 (XML or HDF5 format)
• Graph (see Ex2)

Can be simulated using:

• NetPyNE
• jNeuroML
• NEURON generated from jNeuroML
• NetPyNE generated from jNeuroML

Ex. 4: A network with PyNN cells & inputs

JSON | Python script | Generated NeuroML2

Can be exported to:

• NeuroML 2 (XML or HDF5 format)
• Graph (see above)

Can be simulated using:

• NEST via PyNN
• NEURON via PyNN
• Brian via PyNN
• jNeuroML
• NEURON generated from jNeuroML
• NetPyNE generated from jNeuroML

Ex. 5: A network with the Blue Brain Project connectivity data

JSON | Python script

Can be exported to:

• NeuroML 2 (XML or HDF5 format)
• Graph (see above)
• Matrix (see above)

Can be simulated using:

• NetPyNE

Ex. 6: A network based on Potjans and Diesmann 2014 (work in progress)

JSON | Python script | Generated NeuroML2

Can be exported to:

• NeuroML 2 (XML or HDF5 format)
• Graph (see above)
• Matrix (see above)

Ex. 7: A network based on Brunel 2000 (work in progress)

JSON | Python script | Generated NeuroML2

Can be exported to:

• NeuroML 2 (XML or HDF5 format)
• Graph (see above)

Can be simulated using:

• jNeuroML

Installation & usage

Installation of the basic framework should be fairly straightforward:

git clone https://github.com/NeuroML/NeuroMLlite.git
cd NeuroMLlite
sudo python setup.py install

Then simple examples can be run:

cd examples
python Example1.py  #  Generates the JSON representation of the network (console & save to file)

To generate the NeuroML 2 version of the network, first install pyNeuroML, then use the -nml flag:

sudo pip install pyNeuroML
python Example2.py -nml       # Saves the network structure to a *net.nml XML file

Other options (which will require Neuron, NetPyNE, PyNN, NEST, Brain etc. to be installed) include:

python Example4.py -jnml       # Generate NeuroML2 & LEMS simulation & run using jNeuroML
python Example4.py -jnmlnrn    # Generate NeuroML2 & LEMS simulation, use jNeuroML to generate Neuron code (py/hoc/mod), then run in Neuron
python Example4.py -jnmlnrn    # Generate NeuroML2 & LEMS simulation, use jNeuroML to generate NetPyNE code (py/hoc/mod), then run in NetPyNE
python Example4.py -netpyne    # Generate network in NetPyNE directly & run simulation
python Example4.py -pynnnrn    # Generate network in PyNN, run using simulator Neuron
python Example4.py -pynnnest   # Generate network in PyNN, run using simulator NEST
python Example4.py -pynnbrian  # Generate network in PyNN, run using simulator Brian

Graphs of the network structure can be generated at many levels of detail (1-6) and laid out using GraphViz engines (d - dot (default); c - circo; n - neato; f - fdp). See above images for generated examples.

python Example6.py -graph3d
python Example6.py -graph2f
python Example6.py -graph1n

Other examples

NeuroMLlite is being tested/used in the following repositories on OSB:

• Wilson & Cowan
• Mejias et al. 2016
• Pospischil et al. 2008
• Bezaire et al. 2016
• PING networks
• PsyNeuLink tests
• OpenWorm - c302

See also:

• ModECI MDF