Dear all,
I would like to simulate a network of current based LIF ('iaf_psc_delta') with inhibitory plasticity.
I tried using the model vogels_sprekeler_synapse, but the resulting weights are positive (and they increase when firing rate is high). I guess it is because when used with conductance based neurons, the resulting weights would be multiplied by a gI<0? And with current based neurons, this is not the case. The weight is directly used as the synaptic weight. Do I understand it right?
Is there a way to implement inhibitory plasticity using the vogels_sprekeler_synapse and current based LIF?
Best, Júlia
Dear Júlia,
Indeed, in some neurons that have only one input port, distinguishing excitatory from inhibitory spikes is performed on the basis of the sign of the weight. To avoid this, I would suggest to use a "multisynapse" type neuron model, that is one that has more than one input port. You can thus dedicate one port for inhibitory connections with strictly positive weights (so you can use the vogels_sprekeler_connection) and move the minus sign to the definition of I_syn or to the differential equation for V_m.
Unfortunately there is no multisynapse version of the iaf_psc_delta neuron, but for an example you can look at iaf_psc_exp_multisynapse. If you need delta-function postsynaptic responses, you'll have to modify the exp_multisynapse model a little bit. Please let us know if you run into any trouble.
I'd also like to mention that you can use NESTML to write your neuron models in, so that you don't have to write any C++. You can find more information about NESTML on https://github.com/nest/nestml/. Also here please let us know if you run into any issues!
With kind regards, Charl
On Wed, May 6, 2020, at 12:49, Julia Gallinaro wrote:
Dear all,
I would like to simulate a network of current based LIF ('iaf_psc_delta') with inhibitory plasticity.
I tried using the model vogels_sprekeler_synapse, but the resulting weights are positive (and they increase when firing rate is high). I guess it is because when used with conductance based neurons, the resulting weights would be multiplied by a gI<0? And with current based neurons, this is not the case. The weight is directly used as the synaptic weight. Do I understand it right?
Is there a way to implement inhibitory plasticity using the vogels_sprekeler_synapse and current based LIF?
Best, Júlia _______________________________________________ NEST Users mailing list -- users@nest-simulator.org To unsubscribe send an email to users-leave@nest-simulator.org
Dear Charl,
thank you for your reply! I am not familiar with NESTML yet, but I will go through the documentation you sent and try to implement a model as you suggested. Using exponential PSC would not be a problem for me.
I had meanwhile found a workaround, though. When setting the parameters for the vogels_sprekeler_synapse, I set both 'Wmax' and the parameter 'weight' to a negative value:
nest.CopyModel('vogels_sprekeler_synapse', 'plastic_inhibitory', {'tau': tau_stdp, 'Wmax': -10., 'eta': eta, 'alpha': alpha, 'weight': -.1, 'weight_recorder': weight_recorder[0]})
It now works, and seems to give me the results as expected. Would you know if this is equivalent to using the rule with positive weights and multiplying it by -1 afterwards?
Best, Júlia
On Sat, 09 May 2020 13:08:20 +0200 "Charl Linssen" nest-users@turingbirds.com wrote:
Dear Júlia,
Indeed, in some neurons that have only one input port, distinguishing excitatory from inhibitory spikes is performed on the basis of the sign of the weight. To avoid this, I would suggest to use a "multisynapse" type neuron model, that is one that has more than one input port. You can thus dedicate one port for inhibitory connections with strictly positive weights (so you can use the vogels_sprekeler_connection) and move the minus sign to the definition of I_syn or to the differential equation for V_m.
Unfortunately there is no multisynapse version of the iaf_psc_delta neuron, but for an example you can look at iaf_psc_exp_multisynapse. If you need delta-function postsynaptic responses, you'll have to modify the exp_multisynapse model a little bit. Please let us know if you run into any trouble.
I'd also like to mention that you can use NESTML to write your neuron models in, so that you don't have to write any C++. You can find more information about NESTML on https://github.com/nest/nestml/. Also here please let us know if you run into any issues!
With kind regards, Charl
On Wed, May 6, 2020, at 12:49, Julia Gallinaro wrote:
Dear all,
I would like to simulate a network of current based LIF ('iaf_psc_delta') with inhibitory plasticity.
I tried using the model vogels_sprekeler_synapse, but the resulting weights are positive (and they increase when firing rate is high). I guess it is because when used with conductance based neurons, the resulting weights would be multiplied by a gI<0? And with current
based
neurons, this is not the case. The weight is directly used as the synaptic weight. Do I understand it right?
Is there a way to implement inhibitory plasticity using the vogels_sprekeler_synapse and current based LIF?
Best, Júlia _______________________________________________ NEST Users mailing list -- users@nest-simulator.org To unsubscribe send an email to users-leave@nest-simulator.org
NEST Users mailing list -- users@nest-simulator.org To unsubscribe send an email to users-leave@nest-simulator.org
On Mon, May 11, 2020 10:05:00 +0200, Julia Gallinaro wrote:
Dear Charl,
Hi Julia,
thank you for your reply! I am not familiar with NESTML yet, but I will go through the documentation you sent and try to implement a model as you suggested. Using exponential PSC would not be a problem for me.
I had meanwhile found a workaround, though. When setting the parameters for the vogels_sprekeler_synapse, I set both 'Wmax' and the parameter 'weight' to a negative value:
nest.CopyModel('vogels_sprekeler_synapse', 'plastic_inhibitory', {'tau': tau_stdp, 'Wmax': -10., 'eta': eta, 'alpha': alpha, 'weight': -.1, 'weight_recorder': weight_recorder[0]})
It now works, and seems to give me the results as expected. Would you know if this is equivalent to using the rule with positive weights and multiplying it by -1 afterwards?
The sign of the weight is not changed by the depression/facilitation functions. They only modify the magnitude and then copy the sign back on to the value. So, using negative values for weight and Wmax is the correct thing to do here.
This is the C++ implementation for your reference: https://github.com/nest/nest-simulator/blob/master/models/vogels_sprekeler_c...
We check to ensure that Wmax and the initial weight are of the same sign, so that an inhibitory synapses cannot be flipped to an excitatory and vice-versa: https://github.com/nest/nest-simulator/blob/master/models/vogels_sprekeler_c...
I *think* a negative weight is "standard convention" for inhibitory synapses in NEST, but I cannot find a reference to this at the moment.
On Mon, May 11, 2020 11:44:27 +0100, Ankur Sinha wrote:
We check to ensure that Wmax and the initial weight are of the same sign, so that an inhibitory synapses cannot be flipped to an excitatory and vice-versa: https://github.com/nest/nest-simulator/blob/master/models/vogels_sprekeler_c...
Uh: ".. so that an inhibitory synapse cannot be flipped to an excitatory one and vice versa".
Thank you, Ankur!
I will then assign negative values for weight and Wmax.
Best, Júlia
On Mon, 11 May 2020 11:44:27 +0100 Ankur Sinha sanjay.ankur@gmail.com wrote:
On Mon, May 11, 2020 10:05:00 +0200, Julia Gallinaro wrote:
Dear Charl,
Hi Julia,
thank you for your reply! I am not familiar with NESTML yet, but I
will
go through the documentation you sent and try to implement a model
as
you suggested. Using exponential PSC would not be a problem for me.
I had meanwhile found a workaround, though. When setting the
parameters
for the vogels_sprekeler_synapse, I set both 'Wmax' and the
parameter
'weight' to a negative value:
nest.CopyModel('vogels_sprekeler_synapse', 'plastic_inhibitory', {'tau': tau_stdp, 'Wmax': -10., 'eta': eta, 'alpha': alpha, 'weight': -.1, 'weight_recorder': weight_recorder[0]})
It now works, and seems to give me the results as expected. Would
you
know if this is equivalent to using the rule with positive weights
and
multiplying it by -1 afterwards?
The sign of the weight is not changed by the depression/facilitation functions. They only modify the magnitude and then copy the sign back on to the value. So, using negative values for weight and Wmax is the correct thing to do here.
This is the C++ implementation for your reference: https://github.com/nest/nest-simulator/blob/master/models/vogels_sprekeler_c...
We check to ensure that Wmax and the initial weight are of the same sign, so that an inhibitory synapses cannot be flipped to an excitatory and vice-versa: https://github.com/nest/nest-simulator/blob/master/models/vogels_sprekeler_c...
I *think* a negative weight is "standard convention" for inhibitory synapses in NEST, but I cannot find a reference to this at the moment.
-- Thanks, Regards, Ankur Sinha (He / Him / His) | https://ankursinha.in PhD candidate in computational neuroscience at UH Biocomputation | http://biocomputation.herts.ac.uk/ Research Fellow at the Silver Lab | http://silverlab.org/ Free/Open source community volunteer at the Fedora project | https://fedoraproject.org/wiki/User:Ankursinha Time zone: Europe/London
-
Ankur Sinha -
Charl Linssen -
Julia Gallinaro