Stochastic mech features

[AurelienJaquier]

• new method that either set to deterministic, or set back to stochastic (if stochastic mechanism)
• always set stoch data, even when used in deterministic mode. Reason: some neuron function (e.g. psection) will look up the mechanism data, and if it points to nothing, it crashes

[DrTaDa]

@wvangeit this is something that was implemented for BluePyEModel but we thought it might be better in BluePyOpt

[DrTaDa]

@AurelienJaquier it makes me think, maybe all of the stochastic logic could be brought to BluePyOpt. We should have a look.

[AurelienJaquier]

@DrTaDa Do you mean the logic in the run() method of BPEM_Protocol? Or were you thinking of something else?

[wvangeit]

Should the pre_protocol change the value to deterministic, run it, and set it back to the old value?

[DrTaDa]

@DrTaDa Do you mean the logic in the run() method of BPEM_Protocol? Or were you thinking of something else?

@AurelienJaquier yes, this part.

The only part left in BPEM would be the instantiation of the simulator, stochastic or not. But maybe that part can also be put in the BluePyOpt simulator class.

[AurelienJaquier]

Ok, I'll try both your ideas.

[DrTaDa]

Looks good to me