Animations, and saving them

YBCS improved our plotting code to allow creation and saving of animations, using Matplotlib. This is useful for analysis, and also for dissemination.

So, for instance, one can generate this animation from our single izhikevich neuron documentation example:

An animated GIF showing the membrane potential of a regular spiking Izhikevich neuron.

Here's an example script that reads in the data generated by our example and saves the animation in a GIF file:

#!/usr/bin/env python3
from pyneuroml.plot import generate_plot
from pyneuroml.runners import reload_saved_data

data = reload_saved_data("./LEMS_example-single-izhikevich2007cell-sim.xml")
generate_plot([data['t']], [data['IzhPop0[0]/v']], animate=True,
              save_figure_to="izhikevich.gif", title="Membrane potential")

Generating the animation doesn't take so much time, but saving it does. From what I've managed to find (for example, on this ticket), saving of the animation relies on piping it to external tools ("animation writers"), and that bit seems to take time. We use a progressbar to provide some feedback to the user, so they have some idea of how long to expect it to take.

python test-animate.py
pyNeuroML >>> INFO - Generating plot: Membrane potential
pyNeuroML >>> INFO - Saving animation of 5000ms to izhikevich.gif using pillow
pyNeuroML >>> INFO - This could take a while..
10000 of 10000|#######################################################################################################################################|Elapsed Time: 0:03:29
pyNeuroML >>> INFO - Saved animation to izhikevich.gif
pyNeuroML >>> INFO - Closing plot

Using Vispy visualisations in Jupyter notebooks

Vispy visualisations can also work in Jupyter notebooks using jupyter_rfb. It'll be very good for us to make this work for our visualisations too, so that users can view their cells and networks in 3D in their notebooks. I tinkered with it a bit but could not get jupyter_rfb to work for me. I filed a bug upstream, and it was quickly fixed. I now need to continue playing with it to see what changes, if any, we need to make to our code to make our visualisations work.

Google Summer of Code

We had our first meeting with our two GSoC interns. It's still only the community bonding period, so they're just tinkering with the code and reading the documentation to get a feel for how things work. They've both already been working on pull requests, so they're already ahead of time.

It was really good to have a chat with them. We talked about what they hope to get out of the internship and so on, so that we can tweak the tasks to ensure that the internship benefits them.

Model standardization

I continued to work on converting the Purkinje cell model from Zang et al. I'm still working on the biophysics. There are a few ion channels where the dynamics are described in kinetic state formalism that one has to convert manually. Analysing these to verify that they've been converted correctly is hard too, because unlike ion channels that follow the standard Hodgkin-Huxley formalism, the gates in the Kinetic State ion channels don't have their steady state and decay time constants directly available. So, we're going to have to simulate a passive cell with a single Kinetic State ion channel, both for the original NEURON implementation of the ion channel and for our NeuroML conversion, and then we'll compare the various generated biophysics (and iteratively tweak our conversion until they match).

OSBv2: dev environment

Since OSBv2 is cloud based (currently deployed on Google Cloud), it makes use of containers and Kubernetes (our software partners, MetaCell manage our infrastructure and a majority of the development/maintenance of OSB). The complete platform is also broken down into a number of services. To test changes and develop, therefore, one has to run a local deployment. We've documented the steps in our readme, and I've got a script to automate the various steps.

Of course, it's never that simple. While my script had worked the last time I had tinkered with OSBv2, it didn't work last week and I spent hours trying to figure out why. I need to continue with that this week too.