Connectors Examples

This page provides examples of how to use the Connectors module for building complex network connectivity patterns in BMTK models.

Basic Setup

All connector examples use the following network node structure:

from bmtk.builder import NetworkBuilder

# Create main network
net = NetworkBuilder('example_net')
net.add_nodes(N=100, pop_name='PopA', model_type='biophysical')
net.add_nodes(N=100, pop_name='PopB', model_type='biophysical')

# Create background inputs
background = NetworkBuilder('background')
background.add_nodes(N=300, pop_name='tON', potential='exc', model_type='virtual')

Unidirectional Connector

The UnidirectionConnector allows you to build unidirectional connections with a given probability:

from bmtool.connectors import UnidirectionConnector

# Create connector with 15% connection probability and 1 synapse per connection
connector = UnidirectionConnector(p=0.15, n_syn=1)

# Set up source and target nodes
connector.setup_nodes(source=net.nodes(pop_name='PopA'), target=net.nodes(pop_name='PopB'))

# Add the edges to the network
net.add_edges(**connector.edge_params())

This creates connections from neurons in population 'PopA' to neurons in population 'PopB' with a 15% probability of connection.

Reciprocal Connector

The ReciprocalConnector enables you to build connections with a specific reciprocal probability:

from bmtool.connectors import ReciprocalConnector

# Create connector with 15% base probability and 6.7% reciprocal probability
connector = ReciprocalConnector(
    p0=0.15,           # Base connection probability 
    pr=0.06767705087,  # Reciprocal connection probability
    n_syn0=1,          # Number of synapses for base connection
    n_syn1=1,          # Number of synapses for reciprocal connection
    estimate_rho=False # Whether to estimate rho value
)

# Setup for recurrent connections within PopA
connector.setup_nodes(source=net.nodes(pop_name='PopA'), target=net.nodes(pop_name='PopA'))

# Add the edges to the network
net.add_edges(**connector.edge_params())

This creates connections within population 'PopA' where: - Any two neurons have a 15% probability of having a unidirectional connection - If a unidirectional connection exists from neuron A to neuron B, there's a 6.7% probability of also having a connection from B to A

Correlated Gap Junction

The CorrelatedGapJunction connector creates gap junction connections that can be correlated with chemical synapses:

from bmtool.connectors import ReciprocalConnector, CorrelatedGapJunction

# First create a chemical synapse connectivity pattern
connector = ReciprocalConnector(p0=0.15, pr=0.06, n_syn0=1, n_syn1=1, estimate_rho=False)
connector.setup_nodes(source=net.nodes(pop_name='PopA'), target=net.nodes(pop_name='PopA'))
net.add_edges(**connector.edge_params())

# Then create gap junctions that are correlated with chemical synapses
gap_junc = CorrelatedGapJunction(
    p_non=0.1228,  # Probability for pairs with no chemical synapse
    p_uni=0.56,    # Probability for pairs with unidirectional chemical synapse
    p_rec=1,       # Probability for pairs with reciprocal chemical synapses
    connector=connector  # Use the chemical synapse connector for correlation
)
gap_junc.setup_nodes(source=net.nodes(pop_name='PopA'), target=net.nodes(pop_name='PopA'))

# Add gap junction edges
conn = net.add_edges(
    is_gap_junction=True, 
    syn_weight=0.0000495, 
    target_sections=None,
    afferent_section_id=0, 
    afferent_section_pos=0.5,
    **gap_junc.edge_params()
)

This creates gap junctions with probabilities that depend on the existing chemical synapse connections.

One-to-One Sequential Connector

The OneToOneSequentialConnector creates one-to-one mappings between populations:

from bmtool.connectors import OneToOneSequentialConnector

# Create the connector
connector = OneToOneSequentialConnector()

# Connect background to PopA
connector.setup_nodes(source=background.nodes(), target=net.nodes(pop_name='PopA'))
net.add_edges(**connector.edge_params())

# Connect background to PopB
connector.setup_nodes(target=net.nodes(pop_name='PopB'))
net.add_edges(**connector.edge_params())

This creates one-to-one connections where each neuron in the background population connects to exactly one neuron in PopA and one in PopB.

Advanced Usage

Distance-Dependent Connectivity

You can create distance-dependent connections by providing a connection probability function:

import numpy as np
from bmtool.connectors import UnidirectionConnector

# Define a distance-dependent probability function
def connection_probability(source, target, distance_range=500.0, p_max=0.15):
    """Probability decreases with distance"""
    dist = np.sqrt(np.sum((source['positions'] - target['positions'])**2, axis=1))
    return p_max * np.exp(-dist/distance_range)

# Create connector with the probability function
connector = UnidirectionConnector(p=connection_probability, n_syn=1)
connector.setup_nodes(source=net.nodes(pop_name='PopA'), target=net.nodes(pop_name='PopB'))
net.add_edges(**connector.edge_params())

This creates connections where the probability decreases exponentially with distance between neurons.