sns_toolbox.networks

Networks store neuron and synapse types, and the pattern of connections between them. While not capable of being simulated directly, they are used as a template for compilation.

Classes

Network	Base class for all Networks.
AdditionNetwork	Network which performs addition or subtraction of multiple inputs. Currently only supports non-spiking neurons.
MultiplicationNetwork	Network where the activity in neuron 0 is roughly multiplied by the activity in neuron 1. Currently only supports
DivisionNetwork	Network where the activity in neuron 0 is roughly divided by the activity in neuron 1. Currently only supports
DifferentiatorNetwork	Network where the activity of 'Uout' is the derivative of the activity in 'Uin'. Currently only supports
IntegratorNetwork	Network where the activity of 'Uint' integrates the applied stimulus to itself over time. Currently only supports

Functions

__kernel_connections_1d__(pop_size, kernel[, fill_value])	Generate a connection matrix from a kernel vector and population size.
__kernel_connections_2d__(pop_shape, kernel[, fill_value])	Generate a connection matrix from a kernel matrix and population shape.

Module Contents

class sns_toolbox.networks.Network(name: str = 'Network')

Base class for all Networks.

Parameters:

name (str, optional) – Name for this network, defaults to ‘Network”.

params: Dict[str, Any]

inputs = []

populations = []

outputs = []

outputConns = []

connections = []

get_num_neurons() → int

Calculate the number of neurons in the network, including within populations.

Returns:

The number of neurons in the network.

Return type:

int

get_num_connections() → int

Calculate the number of connections in the network. Does not account for individual synapses between neurons in populations.

Returns:

The number of connections in the network.

Return type:

int

get_num_populations() → int

Get the number of populations in the network.

Returns:

Number of populations.

Return type:

int

get_num_inputs() → int

Get the number of input nodes to the network. Vector inputs are treated as a single input node.

Returns:

Number of inputs.

Return type:

int

get_num_inputs_actual() → int

Calculate the number of individual inputs throughout the network. Vector inputs have one input per dimension.

Returns:

Number of individual inputs.

Return type:

int

get_num_outputs() → int

Get the number of output nodes from the network. Vector outputs have one output node.

Returns:

Number of outputs.

Return type:

int

get_num_outputs_actual() → int

Calculate the number of individual outputs from the network. Vector outputs have one output per dimension.

Returns:

Number of individual outputs.

Return type:

int

get_population_index(name: str) → int

Given a string, find the numerical index of the population corresponding to that name within the network.

Parameters:

name (str) – Name of the population to find.

Returns:

Index of the population.

Return type:

int

get_connection_index(name: str) → int

Given a string, find the numerical index of the connection corresponding to that name within the network.

Parameters:

name (str) – Name of the connection to find.

Returns:

Index of the connection.

Return type:

int

get_input_index(name: str) → int

Given a string, find the numerical index of the input node given by that name within the network.

Parameters:

name (str) – Name of the input node to find.

Returns:

Index of the input node.

Return type:

int

add_population(neuron_type: sns_toolbox.neurons.Neuron, shape, name: str = None, color=None, initial_value=None) → None

Add a neural population to the network.

Parameters:

• neuron_type (sns_toolbox.design.neurons.Neuron) – Type of neuron to add.

• shape (np.ndarray or torch.tensor) – The number of neurons in each dimension of the population.

• name (str, optional) – Name of the population, default is the original name.

• color (str, optional) – Color of the population in the rendered image, default is the original color.

• initial_value (class: 'numbers.Number', class: 'np.ndarray', or class: 'torch.tensor') – Initial value of membrane voltage for each neuron in the population. Must be either a single value, or an array matching ‘shape’. Default value is the original value.

Returns:

None

Return type:

N/A

add_neuron(neuron_type: sns_toolbox.neurons.Neuron, name=None, color=None, initial_value=None) → None

Add a neuron to the network. Note that this is just a special case of addPopulation, which makes a population of 1 neuron.

Parameters:

• neuron_type (sns_toolbox.design.neurons.Neuron) – Type of neuron to add.

• name (str, optional) – Name of the neuron, defaults to the original name.

• color (str, optional) – Color of the neuron in the visual render, defaults to the original color.

• initial_value (Number) – Initial value of the membrane voltage, defaults to 0.0. Units are millivolts (mV).

Returns:

None

Return type:

N/A

add_input(dest: Any, size: int = 1, name: str = 'Input', color='white') → None

Add an input source to the network.

Parameters:

• dest (int or string) – Destination this input connects to.

• size (int, optional) – Number of input elements, default is 1.

• name (str, optional) – Name of the input node, default is ‘Input’.

• color (str, optional) – Color of the input node in the visual render, default is ‘white’.

Returns:

None

Return type:

N/A

add_output(source: Any, name: str = 'Output', spiking: bool = False, color: str = 'white') → None

Add an output node to the network.

Parameters:

• source (int or str) – Source this output is connected to.

• name (str, optional) – Name of the node, defaults to ‘Output’.

• spiking (bool, optional) – Flag for if this node stores voltage or spikes, default is ‘False’.

• color (str, optional) – Color of the output in the visual render, default is ‘white’.

Returns:

None

Return type:

N/A

add_connection(connection_type: sns_toolbox.connections.Connection, source: Any, destination: Any, name: str = None, view_label: bool = False) → None

Add a synaptic connection between two populations in the network.

Parameters:

• connection_type (sns_toolbox.design.connections.Connection) – Type of connection to add.

• source (int or str) – Index or name of source population in the network.

• destination (int or str) – Index or name of destination population in the network.

• name (str, optional) – Name of synapse, default is the original name.

• view_label (bool, optional) – Flag to render the name on the output graph, default is ‘False’.

Returns:

None

Return type:

N/A

add_network(network: Network, color: str = None, suffix: str = None) → None

Add an existing topology of inputs, outputs, and populations to the network.

Parameters:

• network (sns_toolbox.design.networks.Network) – Network to copy over.

• color (str, optional) – Color to render nodes in the network, default is the original colors.

• suffix (str, optional) – String to append to all the population/neuron names in the network.

Returns:

None

Return type:

N/A

copy()

Create a copy of the network.

Returns:

A new network with the same properties as the original. Each can be edited without effecting the other.

Return type:

sns_toolbox.design.networks.Network

compile(dt=0.01, backend='numpy', device='cpu', debug=False, return_params=False) → sns_toolbox.backends.Backend

sns_toolbox.networks.__kernel_connections_1d__(pop_size, kernel, fill_value=0.0)

Generate a connection matrix from a kernel vector and population size.

Parameters:

• pop_size (int) – Number of neurons in the population.

• kernel (list, np.ndarray, or torch.tensor) – Kernel vector to apply.

Returns:

Connection matrix.

Return type:

np.ndarray

sns_toolbox.networks.__kernel_connections_2d__(pop_shape, kernel, fill_value=0.0)

Generate a connection matrix from a kernel matrix and population shape.

Parameters:

• pop_shape (list, np.ndarray, or torch.tensor) – Shape of the population.

• kernel (np.ndarray or torch.Tensor) – Kernel matrix to apply.

Returns:

Connection matrix.

Return type:

np.ndarray

class sns_toolbox.networks.AdditionNetwork(gains, add_del_e=100, sub_del_e=-40, neuron_type=NonSpikingNeuron(), name='Add', R=20.0, **kwargs)

Bases: Network

Network which performs addition or subtraction of multiple inputs. Currently only supports non-spiking neurons.

Parameters:

• gains (list, np.ndarray, or torch.tensor) – List of addition or subtraction weights.

• add_del_e (Number, optional) – Reversal potential of addition synapses, default is 100. Unit is millivolts (mV).

• sub_del_e (Number, optional) – Reversal potential of subtraction synapses, default is -40. Unit is millivolts (mV).

• neuron_type (sns_toolbox.design.neurons.NonSpikingNeuron, optional) – Neuron preset to use, default is sns_toolbox.design.neurons.NonSpikingNeuron.

• name (str, optional) – Name of this network, default is ‘Add’.

num_inputs

class sns_toolbox.networks.MultiplicationNetwork(neuron_type=NonSpikingNeuron(), name='Multiply', R=20.0, **kwargs)

Bases: Network

Network where the activity in neuron 0 is roughly multiplied by the activity in neuron 1. Currently only supports non-spiking neurons.

Parameters:

• neuron_type (sns_toolbox.design.neurons.NonSpikingNeuron, optional) – Neuron preset to use, default is sns_toolbox.design.neurons.NonSpikingNeuron.

• name (str, optional) – Name of this network, default is ‘Multiply’.

transmit

conductance

modulate_special

class sns_toolbox.networks.DivisionNetwork(gain, ratio, name='Divide', neuron_type=NonSpikingNeuron(), R=20.0, **kwargs)

Bases: Network

Network where the activity in neuron 0 is roughly divided by the activity in neuron 1. Currently only supports non-spiking neurons.

Parameters:

• neuron_type (sns_toolbox.design.neurons.NonSpikingNeuron, optional) – Neuron preset to use, default is sns_toolbox.design.neurons.NonSpikingNeuron.

• ratio (Number) – Modulation ratio.

• name (str, optional) – Name of this network, default is ‘Divide’.

transmission

modulation

class sns_toolbox.networks.DifferentiatorNetwork(slew_rate=1.0, name='Differentiate', tau_fast=1.0, R=20.0, **kwargs)

Bases: Network

Network where the activity of ‘Uout’ is the derivative of the activity in ‘Uin’. Currently only supports non-spiking neurons.

Parameters:

• slew_rate (Number, optional) – Steepest signal this differentiator can handle, default is 1.0. Influences the difference in internal time constants.

• name (str, optional) – Name of this network, default is ‘Differentiate’.

• tau_fast (Number, optional) – Time constant of the faster neurons, default is 1.0. Units are milliseconds (ms).

fast_neuron_type

tau_slow

slow_neuron_type

add_synapse

sub_synapse

class sns_toolbox.networks.IntegratorNetwork(integration_gain=0.1, relative_reversal_potential=-40.0, R=20.0, name='Integrator', **kwargs)

Bases: Network

Network where the activity of ‘Uint’ integrates the applied stimulus to itself over time. Currently only supports non-spiking neurons.

Parameters:

• integration_gain (Number, optional) – Gain of the integration, default is 1.0.

• relative_reversal_potential (Number, optional) – Reversal potential of the synapses in the network, default is -40.0. Must be less than 0. Units are millivolts (mV).

• name (str, optional) – Name of this network, default is ‘Integrator’.

membrane_capacitance

neuron_type

synapse_type