gpvolve.paths¶

Path-finding, transition path theory, stochastic sampling.

Pathway finders¶

• shortest_paths(msm, source, targets) -> PathEnsemble (method "shortest")
• greedy_walk(msm, source, targets, *, max_steps=None) -> PathEnsemble (method "greedy")
• dominant_pathways(flux, A, B, *, top_k=10) -> list[PathEnsemble] (method "tpt")

TPT primitives¶

• forward_committor(matrix, A, B) -> NDArray[float64]
• backward_committor(matrix, A, B, *, stationary=None) -> NDArray[float64]
• reactive_flux(matrix, A, B, *, stationary=None) -> csr_matrix
• net_flux(flux) -> csr_matrix
• rate(matrix, A, B, *, stationary=None) -> float

Stochastic sampling¶

@dataclass(frozen=True)
class ConvergenceCheck:
    ess_min: float = 200.0
    rhat_max: float = 1.01
    relevance_threshold: float = 1e-3
    chunk_size: int = 10_000
    max_walkers: int = 1_000_000
    n_chains: int = 4
    max_steps_per_walker: int | None = None

• sample_paths(msm, source, targets, *, convergence=None, seed=None) -> PathEnsemble (method "stochastic"; raises ConvergenceError if max_walkers is exhausted without satisfying both ESS and R-hat thresholds)