Creating a Network

This example walks through how to create a simple network and run it in the software simulator

# sphinx_gallery_thumbnail_path = '_static/example_network.png'

Importing the necessary libraries

First import the CRI network class from the the hs_api library and import the leaky integrate and fire neuron model from the neuron models collection.

from hs_api.api import CRI_network
from hs_api.neuron_models import LIF_neuron

Defining a neuron model

HiAER-Spike supports specifying different models for different neurons. A set of model classes are provided. Currently only variants of the integrate and fire and memoryless integrate and fire model provided by the LIF_neuron class and ANN_neuron class respectively are supported LIF_neuron models have 3 parameters

• theta(\(\theta\)): determines the membrane potential at which the neuron spikes and the potential is reset to zero

• nu(\(\nu\)): controls the magnitude of random noise added to the membrane potential at each step.

• lambda(\(\lambda\)): controls the voltage leakage that occurs during each timestep. \(v=v-v/2^{leak}\)

Neurons a and b belong to neuron model N1, neuron c belongs to neuron model N2, and neuron d belongs to neuron model N3.

N1 = LIF_neuron(theta = 3, nu = -17, Lambda = (2**6)-1)
N2 = LIF_neuron(theta = 4, nu = -17, Lambda = 2)
N3 = ANN_neuron(theta = 5, nu = 2)

Defining the axons dictionary

Axons represent incoming synapses to the network. Each axon has one or more postsynaptic neurons. Users can manually send spikes over axons at each timestep

axons = {'alpha': [('a', 3),('c', 2)],
            'beta': [('b', 3)]}

Defining the neurons dictionary

The neurons dictionary defines the neurons in the network and the synapses between them. Each neuron may have synapses to zero, one, or many postsynaptic neurons and must have a model specified by providing a neuron model object. Keys in the connections and axons dictionaries must be mutually exclusive

neurons = {'a': ([('b', 1), ('d', 2)], N1),
           'b': ([], N1),
           'c': ([], N2),
           'd': ([('c', 1)], N3)}

Defining the outputs List

The outputs list defines the neurons in the network that the user wishes to monitor for spikes. Each element in the list is the key of a neuron in the connections dicitonary

outputs = ['a', 'b']

Initializing a Network

A CRI network object must be contructed using the prevoisly defined dictionaries and list.

network = CRI_network(axons=axons,connections=connections,outputs=outputs)

Running a timestep

The step method executes a single timestep of the network. Inputs may be provided for each timestep in the form of a list of axons to send spikes over dat the given timestep. A list of spikes is always returned. Optionally the membrane potential parameter can be set to true to return membrane potentials for all neurons in the network.

inputs = ['alpha','beta']
currSpikes = network.step(inputs)
# Alternative
# potentials, currSpikes = network.step(inputs, membranePotential=True)
print(currSpikes)

Updating Synapses

Network topologies are frozen at object creation, but the weights of existing synapses may be altered.

currWeight = network.read_synapse('a', 'b')
network.write_synapse('a', 'b', 2)