aimmd.distributed.pathmovers

aimmd.distributed.pathmovers#

This file contains the implementations of various PathMovers.

PathMovers generate new trial paths and accept/reject them, i.e. they move the MarkovChain from one MCStep to the next, or alternatively - depending on if the SP selection scheme supports it - PathMovers can generate new trial paths and directly assign them a weight in the sampled path ensemble without rejecting any trials.

Classes

FixedLengthTwoWayShootingPathMover()

Produce new trials by propagating two trajectories from the selected shooting point, one forward and one backward in time, for a fixed number of frames in total.

PathMover(*[, modelstore, sampler_idx])

Abstract base class defining the API for all PathMover.

RandomVelocitiesShootingPathMover(states, ...)

(Abstract) base class for ShootingPathMover using random velocities at the shooting point.

ShootingPathMover(states, md_engine_spec, *, ...)

Base for all shooting PathMover.

TwoWayShootingPathMover()

Produce new trials by propagating two trajectories from the selected shooting point, one forward and one backward in time, until the given states have been reached.