The reset of after-spike currents in GLIF model does not seem to conform with the original equations about nest-simulator HOT 3 CLOSED

Thanks for your report! I am not an expert on the model, but could it be that the difference you are seeing between the actual and expected values is due to the decay of the after-spike currents during the refractory period? That is, perhaps the variable "ASCurrents_sum" does not just contain the instantaneous update due to the spike, but that plus the amount of decay since the start of the refractory period.

If you set the refractory period to a very small value (0.0001), the actual and expected values match almost exactly.

Does this help?

from nest-simulator.

@clinssen Thanks for the reply! Yes, setting the refractory period to a very small value solves the issue. But it's a little bit strange, as the sum of the after-spike currents is flat during the refractory. It seems like NEST still updating the after-spike currents in the background, yet the actual recorded values are clamped.

I have one more question: Allen institute has release GLIF model parameters fitted to mouse V1 neuron data, and judging from the source code of GLIF neuron in allensdk, it seems that their GLIF model does not silently update after-spike currents during the refractory period. So the behavior of GLIF model in NEST and allensdk is different. It is still valid to use Allen's published GLIF parameters in NEST?

from nest-simulator.

I have checked the code in allensdk further.It turns out that they are also updating the after-spike currents during the refractory period. So the behavior is the same as NEST. Thanks!

from nest-simulator.