Let Neuron SpikeGenerator to create synapses if not present

[filimarc]

At the moment the SpikeGenerator of BSB-NEURON loops over the synapses that are present in the location and stimulates all the selected ones.

If in the SpikeGenerator i select synapses that are not already present they will never be stimulated.

EXAMPLE OF ISSUE

In our neuron example we have:

simulations:
  neuronsim:
    simulator: neuron
    duration: 100
    resolution: 0.025
    temperature: 32
    cell_models:
      stellate_cell:
        model: data.Stellate.definitionStellate
        parameters: []
    connection_models:
      stellate_to_stellate:
        synapses:
          - synapse: GABA
            weight: 0.001
            delay: 1
    devices:
      spike_generator:
        device: spike_generator
        start: 9
        number: 1
        weight: 0.01
        delay: 1
        targetting:
          strategy: by_id
          ids:
            stellate_cell:
              - 0
        locations:
          strategy: branch
          labels:
            - dendrites
        synapses:
          - AMPA
          - NMDA

The only synapse created are GABA, but we want to stimulate through AMPA and NMDA. So in this case the SpikeGenerator do not stimulate anything.

Proposal

I think we can make that the SpikeGenerator check if the synapses specified by the user are present in the location, if not it will create a new synapse.

with small changing in the code:

@config.node
class SpikeGenerator(NeuronDevice, classmap_entry="spike_generator"):
    locations = config.attr(type=LocationTargetting, default={"strategy": "soma"})
    synapses = config.list()
    parameters = config.catch_all(type=types.any_())

    def implement(self, adapter, simulation, simdata):
        for _model, pop in self.targetting.get_targets(
            adapter, simulation, simdata
        ).items():
            for target in pop:
                for location in self.locations.get_locations(target):
                    location_synapses= location.section.synapses
                    names = [syn.name for syn in location_synapses]
                    for synapse in self.synapses:
                        try:
                            index = names.index(synapse)
                            location_synapses[index].stimulate(**self.parameters)
                        except:
                            syn = target.insert_synapse(synapse,location)
                            syn.stimulate(**self.parameters)

[drodarie]

I strongly disagree, hidden features like this would make it so difficult to debug. I would either raise an error or a warning to the user, and not create a synapse. If the user wanted to create this kind of synapses, they should have provided a correct configuration for the neuron models.

[danilobenozzo]

At the moment, the user cannot provide a correct configuration for this use case. Input stimulation is limited to synapses that are defined within a connection_model. I think a solution could be to provide a new spike_generator device able to handle the need to deliver an input through synapses not included in a connection_model.
The main missing information would be the exact locations at which these synapses should be added. Through the location the user can specific on which branches... I don't know if there is way to allow the user to chose on which segments. We may assume that the user specifies the number of synapses, then their segment could be randomly sampled from the specified locations