Connectionists: Linking EEG Functional Connectivity and Excitation–Inhibition Balance to Biophysical Simulation in Alzheimer's Disease on Neuromorphic hardware. GPU-Neuromorphic porting

[Brain and More Lab]

Dear all,

we just published this "Linking EEG Functional Connectivity and
Excitation–Inhibition Balance to Biophysical Simulation in Alzheimer's
Disease" on Neurocomputing:
https://www.sciencedirect.com/science/article/pii/S0925231226022885

Apart the signature of inbalance excitatory/inhibitory in EEG and Nest
simulation.
One of the linking elements was using explainable AI during a
classification with spiking neural networks (SNN). I get to the point of
this mail:

1. We tested spiking neural networks on GPU and on neuromorphic hardware
(Spinnaker 2). Training a model on GPU and then
redeploying it on Spinnaker2
2. For those interested here https://github.com/alecrimi/neurobridge
you can find the porting for the rate-encoding SNN from GPU to Spinnaker.
It is not a library but notebooks.

3. We could not make work the  latency-encoded SNN on Spinnaker. We know
why (see the paper) but if anyone has a solution. Please reach out.

Best,

Szymon, Max and Alex
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/connectionists/attachments/20260901/e03d6e42/attachment.html>