Minimum Action Path (MAP) — FW_minimum_action_path • rareflow

Computes a simple Freidlin–Wentzell minimum action path between two states using gradient descent on the discretized action functional.

Usage

FW_minimum_action_path(
  x0,
  x1,
  drift,
  T = 200,
  dt = 0.01,
  niter = 200,
  stepsize = 0.1
)

Arguments

x0: Starting point (numeric vector).
x1: Target point (numeric vector).
drift: Drift function b(x).
T: Number of time steps.
dt: Time step.
niter: Number of gradient descent iterations.
stepsize: Gradient descent step size.

Value

A list with:

path: matrix of size T × d
action: FW action of the optimized path

Details

This is a lightweight illustrative solver suitable for low-dimensional systems. More advanced solvers (string method, MAM, etc.) can be integrated.