Support for short-term plasticity (TM) on SpiNNaker?

[jazzalin]

Hello,

I am interested in running a simulation with short-term plasticity (mainly TsodyksMarkramSynapses like in pyNN.nest), but it does not seem to be supported currently. I did a number of searches to see if there was any planned support, but only found this 2016 post.

I was wondering if there were any recent developments on that front or if it still outside the scope of releases planned in the near future.

Best regards

[rowleya]

There has been no advancement on this so far. We can take this issue as a request for this to be implemented, but at present can't guarantee a time frame for this.

[jazzalin]

That would be great! In the meantime, I have looked at your tutorials on adding new neuron models and plasticity rules. Out of curiosity, have there been any tutorials/workshops on implementing non-STDP plasticity rules? Thanks!

[rowleya]

I don't think there are any tutorials on that as such, however short-term-plasticity rules would more likely be implemented somewhere in the neuron models. For example, see the "additional input" here: https://github.com/SpiNNakerManchester/sPyNNaker/blob/master/neural_modelling/src/neuron/additional_inputs/additional_input_ca2_adaptive_impl.h

[rowleya]

An alternative might also be to modify the input current, which would follow the interface here: https://github.com/SpiNNakerManchester/sPyNNaker/blob/master/neural_modelling/src/neuron/input_types/input_type_current.h

[rowleya]

If none of those interfaces are suitable, we could of course extend the framework, and would welcome any suggestions in this direction.

[jazzalin]

Thanks for the suggestions, I will have a look.

[dkfellows]

Ref for the Tsodyks Markram model: https://doi.org/10.1073/pnas.94.2.719