Connection Strategies

Connection strategies are rules used to create patterns of connections between source and target groups of neurons. They provide a systematic way to establish synaptic connections with specific properties and patterns. All connection strategies set an excitatory/inhibitory ratio and can randomize weight strengths using probability distributions.

Quick Connection Workflow

The fastest way to create connections is using the 1-2 and 1-3 tricks. See the Building Basics page for a complete explanation.

Tip: After creating connections, the newly created synapses remain selected, allowing you to immediately edit them with Cmd/Ctrl-E, randomize them with R, or open the synapse adjustment panel with Cmd/Ctrl-R.

Connection Methods

Connection strategies can be invoked in several ways:

  • 1-2 and 1-3 tricks: Quick keyboard shortcuts for connecting neurons
  • Right-click menu: Select neurons, right-click, and choose from the “Connect neurons” submenu
  • Synapse groups: Connection strategies are associated with synapse groups for organized connections

Configurable Connection Strategy

When you use the 1-3 trick, Simbrain uses the connection strategy configured in Network Preferences. By default this is All-to-All with 50% excitatory and 50% inhibitory connections.

You can change this strategy in Network Preferences to use any available strategy (Sparse, Distance-Based, Fixed Degree, etc.) with your preferred excitatory/inhibitory ratio.

Two-Stage Connection Process

Connection strategies work in two integrated stages:

  1. Connection Creation: The strategy determines which neurons to connect based on its algorithm (all-to-all, distance-based, sparse, etc.)
  2. Weight Initialization: A weight initializer sets the initial strength values for the created connections

This separation allows you to independently control the topology (which neurons connect) and the weight values (how strong those connections are).

Weight Initializers

After a connection strategy creates synapses, a weight initializer sets their strengths. Three types are available:

Constant

Sets all excitatory synapses to a fixed positive value and all inhibitory synapses to a fixed negative value.

  • Excitatory Strength: The strength value for all excitatory synapses (default 1.0)
  • Inhibitory Strength: The strength value for all inhibitory synapses (default -1.0)

Random

Initializes synapse weights using probability distributions, allowing for variability in connection strengths.

  • Randomize Excitatory: Whether to randomize excitatory connection strengths. When disabled, uses the default excitatory strength (1.0)
  • Randomize Inhibitory: Whether to randomize inhibitory connection strengths. When disabled, uses the default inhibitory strength (-1.0)
  • Excitatory Randomizer: The probability distribution used to generate excitatory weight values
  • Inhibitory Randomizer: The probability distribution used to generate inhibitory weight values
When using randomizers, excitatory/inhibitory polarity is enforced by taking the absolute value of each sampled value. Excitatory weights are set to x and inhibitory weights to - x , where x is the sampled value. This transformation means the observed distribution of weight strengths will differ from the configured distribution, often significantly. For example, a Normal(0, 1) distribution becomes a half-normal distribution. To get expected results, use distributions that naturally produce positive values (like Log-Normal) or adjust your distribution parameters to account for this transformation.

Distance-Based

Scales synapse weights based on the spatial distance between source and target neurons. Closer neurons receive stronger connections.

  • Decay Function: The decay function that determines how weight strength decreases with distance
  • Base Strength: The maximum weight strength at the peak distance (typically distance 0)

For each synapse, the final weight is calculated as: weight = baseStrength × decayFunction(distance). Common decay functions include Gaussian (smooth drop-off), Linear (linear decrease), and Step (constant within radius). Use distance-based initialization when modeling spatially organized networks where nearby neurons should have stronger influence.

Excitatory/Inhibitory Ratio

All connection strategies set the percentage of connections that should be excitatory vs inhibitory. This setting only affects nonpolar neurons (the default). Excitatory and inhibitory neurons will always maintain their polarity constraint regardless of this setting.

Every connection strategy automatically respects the polarity of source neurons:

  • Synapses from excitatory neurons are always positive
  • Synapses from inhibitory neurons are always negative
  • Synapses from nonpolar neurons (default) follow the excitatory/inhibitory ratio

When all source neurons are nonpolar (the default), the ratio straightforwardly sets what percentage of weights will be positive (excitatory) vs negative (inhibitory). Most workflows use nonpolar neurons, where the excitatory/inhibitory ratio setting controls the balance of positive and negative weight signs.

When source neurons have explicit excitatory or inhibitory polarity, the strategy attempts to achieve the requested ratio while respecting individual neuron polarity constraints. Excitatory and inhibitory neurons always maintain their sign constraint, and the ratio is achieved using only nonpolar neurons. For example, if you have 5 excitatory neurons, 5 inhibitory neurons, and 10 nonpolar neurons, and set the ratio to 60% excitatory, the 5 excitatory neurons contribute 5 positive connections, the 5 inhibitory neurons contribute 5 negative connections, and the 10 nonpolar neurons are split 7 excitatory, 3 inhibitory to achieve 12/20 = 60% overall.

Handling Duplicate Connections

Connection strategies handle existing connections differently depending on the context:

Free Weights: When applying a connection strategy to free neurons, existing connections are not overwritten. The strategy creates new connections without checking for duplicates. If you want to replace existing connections, you must manually delete them first. This gives you control to incrementally add connections and fine-tune your network.

Synapse Groups: When applying a connection strategy to a synapse group, all existing synapses in the group are deleted first, and the strategy is applied fresh. This ensures the synapse group always reflects the current strategy settings.

Sparse Strategy (Special Case): The Sparse strategy is designed to allow incremental adjustments to connection density while preserving existing weights:

  • Free weights: Sparse can add new connections or remove existing ones as needed to reach the target density. This is the only strategy that may remove existing free weights.
  • Synapse groups: Sparse maintains awareness of current connections and adjusts them incrementally rather than replacing everything. When increasing density, it adds only the additional connections needed. When decreasing density, it removes connections to reach the target.

This special behavior makes Sparse ideal for iteratively tuning network connectivity without losing carefully configured weights.

Table of contents