DifferenceCoupling

experimental.network_dynamics.coupling.DifferenceCoupling(
    source=None,
    local=None,
    *,
    incoming_states=None,
    local_states=None,
    **kwargs,
)

Diffusive coupling based on state differences.

Computes coupling based on the difference between incoming and local states, implementing:

\[c_i = G \cdot \sum_{j} w_{ij} (x_j - x_i)\]

This type of coupling is useful for synchronization and consensus dynamics, as it drives nodes toward common states. Automatically uses sparse-optimized computation for sparse graphs.

Parameters

Name	Type	Description	Default
source	str or list of str	State name(s) to collect from connected nodes (required)	None
local	str or list of str	State name(s) from current node (required for computing differences)	None

Attributes

Name	Type	Description
N_OUTPUT_STATES	int	Number of output coupling states: 1
DEFAULT_PARAMS	Bunch	Default parameters: G=1.0 (global coupling strength)

Examples

>>> # Diffusive coupling via 'x' state
>>> coupling = DifferenceCoupling(source='x', local='x', G=1.0)

Methods

Name	Description
post	Apply coupling strength to summed differences.
pre	Compute an elementwise difference on aligned messages.

post

experimental.network_dynamics.coupling.DifferenceCoupling.post(
    summed_inputs,
    local_states,
    params,
)

Apply coupling strength to summed differences.

Args: summed_inputs: Summed differences [n_inputs, n_nodes] local_states: Local states (not used) params: Bunch with G

Returns: Scaled coupling [n_inputs, n_nodes]

pre

experimental.network_dynamics.coupling.DifferenceCoupling.pre(
    incoming_states,
    local_states,
    params,
)

Compute an elementwise difference on aligned messages.

Args: incoming_states: Source states [n_incoming, *M]. local_states: Target states aligned as [n_local, *M]. params: Coupling parameters (not used in pre)

Returns: State differences [n_output, *M]. M is a dense target/source grid or the prepared sparse edge axis.