Setting population v after a run gives InvalidParameterType exception

[ej159]

I'm trying to set the voltage of population to a value after a run. It is suggested that this should work on the SpiNNaker Users Group. I receive the following exceptions:

spynnaker.pyNN.exceptions.InvalidParameterType: Population IF_curr_exp does not have parameter v

It then tries to do an intialize() instead but falls over there with this exception:

Exception: initialize can only be called before the first call to run, or before the first call to run after a reset

Here is an example script:

import pyNN.spiNNaker as sim

sim.setup()

pop = sim.Population(1, sim.IF_curr_exp)

sim.run(2)
pop.set(v=0.0)

sim.run(2)

[Andrew Gait]

This shouldn't work with standard PyNN functions, so I think the advice in that message on the users group is somewhat out of date. The argument from the PyNN side as I understand it is that if you want to change the voltage mid-run then you should by definition have to call reset() as well. Is there a reason why you can't call reset() at this point? (Also: the voltage is not a parameter, it's a state variable, so you shouldn't really be calling set(v=..) in the first place).

[I can say that we do also have a non-standard PyNN function pop.set_initial_value(...) which I think is intended to do what you're requesting, but it doesn't work properly for me at the moment; I can work on fixing this if you need it for what you're doing.]

[ej159]

I am using a variable rate Poisson spike source. The way I understand this is that I if I were to call reset() it would reset the simulation time and so begin the simulation from t=0, resetting the VRPSS too. I could reset() and then supply new parameters to the Poisson source. The reason that I didn't do this initially was because reset() was unreliable (I believe this has been improved). I imagine that setting v=0 after each run would be quicker than uploading the rate variables after each run but probably not by much?

[Andrew Gait]

You're correct that reset() has recently been improved so it should work better. I think you're right in terms of setting v=0 being quicker as well but as you say I don't think it would make that much difference. Certainly as long as the change of the rate variables doesn't affect the graph, it shouldn't make that much difference.

[Andrew Gait]

If you use the fix in #332 and SpiNNakerManchester/sPyNNaker#765 then you should now be able to do pop.set_initial_value(variable="v", value=0) in your example script.

[ej159]

I have just found out this problem still exists. When running the example below, the voltage does not get reset to -65.0mV at 50ms as expected (see the output graph). If I try and use Population.set() I get the same InvalidParameterType and initialize can only be called before the first call to run exceptions as I saw before. Is it possible to make set()s between runs work?

import spynnaker8 as p
from pyNN.utility.plotting import Figure, Panel
import matplotlib.pyplot as plt

p.setup(time_scale_factor=1, timestep=1)

pop1 = p.Population(1, p.SpikeSourceArray([10, 100]))
pop2 = p.Population(1, p.IF_curr_exp())

p.Projection(pop1, pop2, p.AllToAllConnector(), p.StaticSynapse(weight=1.0, delay=1.0))

pop2[0].record(['v'])

p.run(50)
pop2.set_initial_value(variable="v", value=-65) #The line in question that falls over with pop2.set(variable="v", value=-65.0))
p.run(200)

v = pop2.get_data('v')

p.end()

Figure(
    # membrane potential of the postsynaptic neuron
    Panel(v.segments[0].filter(name='v')[0],
          ylabel="Membrane potential (mV)",
          data_labels=[pop2.label], yticks=True, xlim=(0, 250), xticks=True),
    annotations="Simulated with {}".format(p.name())
)

plt.show()``` 

[Andrew Gait]

Confirming this is an issue - fix on the way.