Testing/validating NeuroML Models#
Models described in NeuroML can be run across multiple simulators, and it it essential that the activity (e.g. spike times) of the models are as close as possible across all of these independently developed platforms.
It is also important to validate that the behaviour of a given NeuroML model reproduces some recorded aspect of the biological equivalent.
Testing behaviour of NeuroML models across simulators#
This type of testing addresses the question: Does a given NeuroML model produce the same results when run across multiple simulators?
OMV - Open Source Brain Model Validation framework#
The OSB Model Validation framework was originally developed as an automated model validation package for Open Source Brain projects, which can be used for testing model behaviour across many simulation engines both:
on your local machine when developing models
on GitHub Actions, to ensure tests pass on every commit.
See OpenSourceBrain/osb-model-validation for more details.
Validating that NeuroML model reproduce biological activity#
This type of testing addresses the question: How well does a given NeuroML model replicate the activity as seen in real neurons/channels/networks?
SciUnit is a Python framework for test-driven validation of scientific models, and NeuronUnit is a package based on this for data-driven validation of neuron and ion channel models. See also SciDash for more information.
Interactive Jupyter notebooks for running NeuronUnit examples can be found this repository.