gillespie

adapters.gillespie

Gillespie SSA backend — a finite-size stochastic realization of a mean-field rate model.

Runs a relaxation-type rate model as a finite birth-death process (Gillespie 1977). The model must have one activity state variable X obeying a relaxation equation tau*X' = -X + F(state) (a leak -X toward a gain F); any remaining state variables are treated as slow internal variables that evolve deterministically between events. The activity becomes a discrete count n ≈ Omega*X where Omega is the van Kampen system size (execution.system_size): the number of discrete units per unit of X. The rate equation is read as

birth propensity  a+ = Omega * F / tau        (the gain term)
death propensity  a- = n / tau                (the leak term, since a- = Omega*X/tau)

and the slow variables integrate deterministically over each inter-event interval. Finite Omega is the sole source of noise; the deterministic mean field is recovered as Omega -> infinity. Applicable to any single-activity Wilson-Cowan / Tsodyks-Markram type rate model — the birth/death split and the between-event ODEs are derived from the model’s own equations, so nothing here is model-specific.

Reference: Cortes et al. (2013) PNAS 110(41):16610, SI §2 (Eq. S10/S11) and Fig 5.

Classes

Name	Description
GillespieAdapter	Run a mean-field rate `SimulationExperiment` as a finite-N birth-death process.

GillespieAdapter

adapters.gillespie.GillespieAdapter(experiment)

Run a mean-field rate SimulationExperiment as a finite-N birth-death process.

Methods

Name	Description
run	Integrate the experiment with the Gillespie SSA and return the trajectory as an :class:`ExperimentResult`.

run

adapters.gillespie.GillespieAdapter.run(**kwargs)

Integrate the experiment with the Gillespie SSA and return the trajectory as an :class:ExperimentResult.