HOC kernel for Jupyter

[bll5z6]

Hey all,

Would there be any interest in making a HOC kernel so that HOC code can be interpreted in a Jupyter notebook/ the new Jupyter Lab? I'm using it beautifully with NEURON as a Python module but it would be cool to do HOC as well (mainly for teaching). Just thought I'd throw it out there!

Ben

[mgiugliano]

Yes, it would be a terrific idea and similarly it would be nice to have the graphs appear inline.

[ramcdougal]

If you're working on a HOC kernel, I also encourage this...

But:

(1) Consider teaching in Python instead of HOC. That's what we do for the official NEURON courses. The Python programmer's reference is not only the default version, but it's also more up-to-date and has more examples.

(2) As long as you are using Python, then you can embed interactive equivalents of most graphs you might like to do.

(2a) For example, PlotShape (needs NEURON 7.6.1+; download the latest from http://neuron.yale.edu):

Code: Select all

from neuron import h, gui2
gui2.set_backend('jupyter')
h.load_file('geo5038804.hoc')  # or whatever morphology you like
ps = gui2.PlotShape()
ps.variable('v')
ps.show(0)

(Obviously, in the figure there is code just off screen that leads to non-uniform membrane potential.)

The PlotShape is interactive (pan, zoom, 3D rotate) and continuously updates (as you'd expect it to) as variables change.

(2b) Plotting time series with bokeh:

Code: Select all

from neuron import h

# using bokeh for graphics
from bokeh.io import output_notebook
import bokeh.plotting as plt
output_notebook()

# standard run library (gives h.continuerun)
h.load_file('stdrun.hoc')

# morphology
soma = h.Section(name='soma')
soma.L = soma.diam = 20

# biophysics
soma.insert('hh')

# current injection
ic = h.IClamp(soma(0.5))
ic.delay = 1
ic.dur = 0.1
ic.amp = 1

# recording
t = h.Vector()
v = h.Vector()
t.record(h._ref_t)
v.record(soma(0.5)._ref_v)

# initialize and run
h.finitialize(-65)
h.continuerun(10)

# now do the plot
f = plt.figure()
f.line(t, v, line_width=5)
plt.show(f)

(2c) Plotting RangeVarPlot (Space Plots) with bokeh:

Code: Select all

from neuron import h

# using bokeh for graphics
from bokeh.io import output_notebook
import bokeh.plotting as plt
output_notebook()

# standard run library (gives h.continuerun)
h.load_file('stdrun.hoc')

# morphology
axon = h.Section(name='axon')
axon.L = 10000
axon.diam = 1
axon.nseg = 101

# biophysics
axon.insert('hh')

# current injection
ic = h.IClamp(axon(0))
ic.delay = 1
ic.dur = 0.1
ic.amp = 1

# setup a plot

# setup rangevarplot
rvp = h.RangeVarPlot('v')
rvp.begin(axon(0))
rvp.end(axon(1))
x = h.Vector()
y = h.Vector()
ys = []

def save_rvp_data():
    rvp.to_vector(y, x)
    ys.append(list(y))

# initialize and run, plotting twice
h.finitialize(-65)
h.continuerun(5)
save_rvp_data()
h.continuerun(10)
save_rvp_data()

# show the plot
f = plt.figure(x_axis_label='position', y_axis_label='v')
f.multi_line([list(x)] * len(ys), ys, color=['black', 'red'], line_width=2)
plt.show(f)

* The main** known issue with the current Jupyter PlotShape is that it won't re-appear when you reopen a saved notebook; you'll have to rerun the code.
** An h.PlotShape can actually do a lot of things; gui2.PlotShape in 7.6.1 does not reproduce all of the API yet, but it does do a lot of it.

[bll5z6]

This is awesome, thanks for the reply!