Comparison of STDP_dopamine_synapse with brian2

[ferqui]

Hi, I was comparing the NEST stdp_dopamine_synapse with a version in Brian2 simulator, but I get very different results on the eligibility trace.

The equation that I used in the Brian2 implementation is the following:

rstdp_eq = """
dn/dt = -n/tau_n : 1 (clock-driven)
dc/dt = -c/tau_c : 1 (clock-driven)
dweight/dt = c*(n-b) / (1*ms) : 1 (clock-driven)
dApre/dt = -Apre/tau_plus : 1 (clock-driven)
dApost/dt = -Apost/tau_minus : 1 (clock-driven)
b : 1
A_plus : 1
A_minus : 1
tau_n : second
tau_c : second
tau_plus : second
tau_minus : second
"""

rstdp_pre_eq = """
x_post += weight*pA
c += Apost
Apre += A_plus
"""

rstdp_post_eq = """
c += Apre
Apost += A_minus
"""

I am using NEST 2.20.1 and Brian2 2.4.2. Do you have any idea where I might have gone wrong? I attached also the test files that I made. Thank you.
DA-STDP.zip

[clinssen]

Just an initial observation: in Brian, "delay" refers by default to axonal delay (time it takes the spike to arrive from the presynaptic partner to the synapse), whereas in NEST, delay refers to dendritic delay (time it takes for the postsynaptic spike to backpropagate into the dendrite and to the synapse).

Additionally, the synaptic trace value in NEST is only updated when a presynaptic spike is processed. Values obtained inbetween using GetStatus() or get() are cached values, and may not correspond to what the value actually ought to be at that point in time.

[heplesser]

@YounesBouhadjar, could you take a look at this in the context of #820.

[YounesBouhadjar]

I suspect the issue is related to what @clinssen mentioned. We are currently working on testing this synapse model and we will come back to you when we have more insights.

[github-actions]

Issue automatically marked stale!

[heplesser]

@YounesBouhadjar Have you made progress on this?

[clinssen]

@ferqui: could you tell us if there are any updates on this? Otherwise we will close the issue due to the differences in the meaning of "synaptic delay" between simulators (assumed axonal in Brian, and dendritic in NEST) probably being the cause behind the differences.

[github-actions]

Issue automatically marked stale!