braintools.conn module

`braintools.conn` module#

Modular Connectivity System for Neural Network Generation.

This module provides a comprehensive, modular system for building connectivity patterns across different types of neural models. The system is designed with complete decoupling between model types to ensure clean, specialized implementations.

Supported Model Types:

Point Neurons: Single-compartment integrate-and-fire models
Multi-Compartment Models: Detailed morphological neuron models

Key Features:

Direct Class Access: All connectivity patterns available as classes
Biological Realism: Realistic parameters and constraints for each model type
Spatial Awareness: Position-dependent connectivity with proper units
Composable Patterns: Combine and transform connectivity patterns
Extensible Design: Easy to add custom patterns for any model type

Quick Start:

import brainunit as u
from braintools.conn import Random, ExcitatoryInhibitory, AxonToDendrite

# Point neuron random connectivity
random_conn = Random(prob=0.1)
result = random_conn(pre_size=1000, post_size=1000)

# E-I network dynamics
ei_conn = ExcitatoryInhibitory(
    exc_ratio=0.8,
    exc_prob=0.1,
    inh_prob=0.2,
    exc_weight=1.0 * u.nS,
    inh_weight=-0.8 * u.nS
)
result = ei_conn(pre_size=1000, post_size=1000)

# Multi-compartment axon-to-dendrite connectivity
axon_dend = AxonToDendrite(
    connection_prob=0.1,
    weight_distribution='lognormal',
    weight_params={'mean': 2.0 * u.nS, 'sigma': 0.5}
)
result = axon_dend(pre_size=100, post_size=100)

Point Neuron Connectivity:

import numpy as np
import brainunit as u
from braintools.conn import Random, DistanceDependent, ExcitatoryInhibitory

# Realistic synaptic connectivity with proper units
from braintools.init import LogNormal, Normal
ampa_conn = Random(
    prob=0.05,
    weight=LogNormal(mean=1.0 * u.nS, sigma=0.5),
    delay=Normal(mean=1.5 * u.ms, std=0.3 * u.ms)
)

# Spatial connectivity
positions = np.random.uniform(0, 1000, (500, 2)) * u.um
spatial_conn = DistanceDependent(
    sigma=100 * u.um,
    decay='gaussian',
    max_prob=0.3
)
result = spatial_conn(500, 500, positions, positions)

# E-I network with Dale's principle
ei_network = ExcitatoryInhibitory(
    exc_ratio=0.8,
    exc_prob=0.1,
    inh_prob=0.2,
    exc_weight=1.0 * u.nS,
    inh_weight=-0.8 * u.nS
)

Modular connectivity system for building neural network connection patterns across different types of neural models. The system provides specialized implementations for point neurons and multi-compartment models with direct class access.

Base Classes and Results#

Core infrastructure for connectivity patterns and results.

`ConnectionResult`	Universal container for connectivity results across all neuron model types.
`Connectivity`	Abstract base class for all connectivity patterns.
`PointConnectivity`	Base class for point neuron connectivity patterns.
`MultiCompartmentConnectivity`	Base class for multi-compartment neuron connectivity patterns.
`ScaledConnectivity`	Connectivity with scaled weights or delays.
`CompositeConnectivity`	Composite connectivity created by combining patterns.

Point Neuron Connectivity#

Connectivity patterns for single-compartment point neuron models.

Basic Patterns#

Simple connectivity patterns including random and deterministic connections.

`Random`	Random connectivity with fixed connection probability.
`FixedProb`	Alias for Random connectivity with fixed probability (`Random`).
`AllToAll`	Fully connected network where every neuron connects to every other neuron.
`OneToOne`	One-to-one connectivity where neuron i connects to neuron i.

Spatial Patterns#

Distance-dependent and spatially-structured connectivity patterns.

`DistanceDependent`	Distance-dependent connectivity for spatially arranged point neurons.
`Gaussian`	Gaussian distance-dependent connectivity for spatially organized neural networks.
`Exponential`	Exponential distance-dependent connectivity for spatially organized neural networks.
`Ring`	Ring topology connectivity where neurons are arranged in a circular structure.
`Grid2d`	Two-dimensional grid connectivity for spatially arranged neural populations.
`RadialPatches`	Radial patch connectivity where neurons connect within multiple localized spatial patches.
`ClusteredRandom`	Random connectivity with spatial clustering and distance-dependent connection enhancement.

Topological Patterns#

Complex network topology patterns based on graph theory.

`SmallWorld`	Watts-Strogatz small-world network topology.
`ScaleFree`	Barabási-Albert scale-free network with preferential attachment.
`Regular`	Regular network where all neurons have the same degree.
`ModularRandom`	Modular network with intra-module and inter-module random connectivity.
`ModularGeneral`	Modular network using Connectivity instances for both intra and inter-module patterns.
`HierarchicalRandom`	Hierarchical network with multiple levels and asymmetric feedforward/feedback connections.
`CorePeripheryRandom`	Core-periphery network with densely connected core and sparse periphery.

Biological Patterns#

Biologically-inspired connectivity patterns following neural principles.

ExcitatoryInhibitory

Standard excitatory-inhibitory network following Dale's principle.

Kernel-Based Connectivity#

Connectivity patterns using convolution kernels for spatial receptive fields.

`Conv2dKernel`	Convolutional kernel connectivity for spatially arranged point neurons.
`GaussianKernel`	Gaussian kernel connectivity for center-surround receptive fields.
`GaborKernel`	Gabor kernel connectivity for orientation-selective receptive fields.
`DoGKernel`	Difference of Gaussians (DoG) kernel for center-surround receptive fields.
`MexicanHat`	Mexican hat (Laplacian of Gaussian) connectivity pattern.
`SobelKernel`	Sobel edge detection kernel for orientation-selective connectivity.
`LaplacianKernel`	Laplacian kernel for edge detection connectivity.
`CustomKernel`	Custom kernel connectivity using user-defined kernel function.

Multi-Compartment Connectivity#

Connectivity patterns for detailed multi-compartment neuron models with compartment-specific targeting.

Compartment Constants#

Predefined constants for identifying neural compartments.

`SOMA`	int([x]) -> integer int(x, base=10) -> integer
`BASAL_DENDRITE`	int([x]) -> integer int(x, base=10) -> integer
`APICAL_DENDRITE`	int([x]) -> integer int(x, base=10) -> integer
`AXON`	int([x]) -> integer int(x, base=10) -> integer

Basic Compartment Patterns#

Fundamental patterns for compartment-specific connectivity.

`CompartmentSpecific`	General compartment-specific connectivity pattern.
`AllToAllCompartments`	All-to-all compartment connectivity.
`CustomCompartment`	Custom compartment connectivity using user-defined function.

Anatomical Targeting Patterns#

Connectivity patterns based on anatomical organization.

`SomaToDendrite`	Specialized connectivity from soma to dendritic compartments.
`AxonToSoma`	Specialized connectivity from axon to soma compartments.
`DendriteToSoma`	Specialized connectivity from dendrites to soma.
`AxonToDendrite`	Specialized connectivity from axon to dendritic compartments.
`DendriteToDendrite`	Dendrite-to-dendrite connectivity patterns.

Morphology-Aware Patterns#

Patterns that utilize detailed morphological information.

`ProximalTargeting`	Connectivity targeting proximal dendritic compartments.
`DistalTargeting`	Connectivity targeting distal dendritic compartments.
`BranchSpecific`	Branch-specific dendritic targeting.
`MorphologyDistance`	Distance-dependent connectivity based on detailed morphology.

Dendritic Patterns#

Specialized patterns for dendritic targeting and integration.

`DendriticTree`	Dendritic tree connectivity patterns with branch-specific targeting.
`BasalDendriteTargeting`	Specific targeting of basal dendrites.
`ApicalDendriteTargeting`	Specific targeting of apical dendrites.
`DendriticIntegration`	Dendritic integration connectivity patterns.

Axonal Patterns#

Patterns for axonal projection and arborization.

`AxonalProjection`	Axonal projection patterns with topographic organization.
`AxonalBranching`	Axonal branching patterns.
`AxonalArborization`	Axonal arborization patterns.
`TopographicProjection`	Topographic projection patterns.

Synaptic Patterns#

Patterns for synaptic placement and organization.

`SynapticPlacement`	Synaptic placement rules.
`SynapticClustering`	Synaptic clustering patterns.