Two sets of parameters for initializing membrane potential

[heplesser]

The default parameters file defines parameters for initializing the membrane potential of neurons in two places. First we have

multi-area-model/multiarea_model/default_params.py

Lines 78 to 86 in 7110623

	sim_params.update(
	{
	'initial_state': {
	# mean of initial membrane potential (in mV)
	'V_m_mean': -58.0,
	# std of initial membrane potential (in mV)
	'V_m_std': 10.0
	}
	})

and then

multi-area-model/multiarea_model/default_params.py

Lines 107 to 115 in 7110623

	neuron_params = {
	# neuron model
	'neuron_model': 'iaf_psc_exp',
	# neuron parameters
	'single_neuron_dict': single_neuron_dict,
	# Mean and standard deviation for the
	# distribution of initial membrane potentials
	'V0_mean': -100.,
	'V0_sd': 50.}

The comments in both places are very similar and it is not clear to me which of the two sets of parameters are actually used (from

multi-area-model/multiarea_model/simulation.py

Line 468 in 7110623

if self.network.params['USING_NEST_3']:

it seems the second one; the first seems entirely unused).

I find the second set of values rather unphysiological with a mean well below typical GABA reversal potential (-90 mV if I am not wrong), and with one std deviation stretching to -150 mV.

With a spike threshold of -50 mV (

multi-area-model/multiarea_model/default_params.py

Line 93 in 7110623

'V_th': -50.0,

), for the first set of parameters (-58±10 mV) about 21% of all neurons will be initialized to superthreshold membrane potentials, for the second set (-100±50 mV) about 16%. These neurons will all spike simultaneously in the first time step. How sensible is this?

At least for benchmarking, a massive barrage of spikes in a single time step can lead to some problematic (because atypical) resizing of data structures. Since the membrane potential distribution must be zero at the threshold, would it not make more sense to initialize the membrane potential with a lognormal distributions stretching from V_th down towards -inf?