Neural Astrocytic Network

This page demonstrates how a simple Neural-Astrocytic Network (NAN) module can be built on Loihi using combra_loihi.

Feedforward NAN

We build a NAN module by introducing a basic feedforward SNN with 1 input and 1 output layer, a single astrocyte. The astrocyte receives inputs from all input neurons and gives output (SIC) to all output neurons. Input layer and output layer are randomly connected.

import nxsdk.api.n2a as nx
import combra_loihi.api as combra

net = nx.NxNet()
nan = combra.FeedforwardNAN(net)
post_probes, astro_probes = nan.probeNAN([nx.ProbeParameter.SPIKE],
                                         [combra.ASTRO_SPIKE_RECEIVER_PROBE.COMPARTMENT_VOLTAGE,
                                          combra.ASTRO_IP3_INTEGRATOR_PROBE.COMPARTMENT_VOLTAGE,
                                          combra.ASTRO_SIC_GENERATOR_PROBE.COMPARTMENT_VOLTAGE,
                                          combra.ASTRO_SPIKE_GENERATOR_PROBE.SPIKE,
                                          combra.ASTRO_IP3_INTEGRATOR_PROBE.SPIKE])

net.run(sim_time)
net.disconnect()

A more detailed example of a Feedforward NAN module can be found here.

Feedforward NAN with astrocyte-induced synchronous activity