-
Notifications
You must be signed in to change notification settings - Fork 5
Feedforward NAN
A basic feedforward spiking neural network with 1 input layer, 1 output layer, and 1 Astrocyte.
There are some configurable parameters such as the current and voltage decay for the Loihi compartments representing the neurons.
Number of pre-synaptic neurons.
(Default=20)
Number of post-synaptic neurons.
(Default=20)
Average firing rate for the pre-synaptic neurons.
(Default=20)
Weight between pre-synaptic neurons and post-synaptic neurons.
(Default=20)
Connection density
(Default=0.1)
Post-synaptic neurons voltage threshold. The actual threshold for compartmentVoltage is 2^6 * post_vth.
(Default=100)
compartmentCurrentDecay for post-synaptic neuron compartments.
(Default=409)
compartmentVoltageDecay for post-synaptic neuron compartments.
(Default=1024)
Simulation time duration in milliseconds.
(Default=30000)
Parameters:
-
postConditionsThe list of Loihi compartment node parameters to probe. -
astroConditionsEither one value or a list of values indicating which property and which compartment to probe within the Astrocyte.All possible enum values are located under
combra.ASTRO_SPIKE_RECEIVER_PROBE,combra.ASTRO_IP3_INTEGRATOR_PROBE,combra.ASTRO_SIC_GENERATOR_PROBE,combra.ASTRO_SPIKE_GENERATOR_PROBEwith each compartment's properties as values within that probe class respectively.
import combra_loihi.api as combra
nan = combra.FeedforwardNAN(net, sim_time=sim_time)
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])Guangzhi Tang, Arpit Shah, Computational Brain Lab, Computer Science Department, Rutgers University