wepy.runners.randomwalk module

The random walk dynamics runner.

In this system, the state of the walkers is defined as an N-dimensional vector of non-negative values. The walkers start at position zero (in N-dimensional space) and randomly move a step either forward or backward with the given probabilities. This is done in each dimension at each dynamic step. All moves that result in a negative position are rejected.

One potentioanl use of the random walk system is to test the performance of differnt resamplers as seen in these papers:

“WExplore: Hierarchical Exploration of High-Dimensional Spaces Using the Weighted Ensemble Algorithm” and “REVO: Resampling of Ensembles by Variation Optimization”.

wepy.runners.randomwalk.UNIT_NAMES = (('positions_unit', 'microsecond'), ('time_unit', 'picosecond'))

Mapping of units identifiers to the corresponding pint units.

class wepy.runners.randomwalk.RandomWalkRunner(probability=0.25)

Bases: wepy.runners.runner.Runner

RandomWalk runner for random walk simulations.

Constructor for RandomWalkRunner.

Parameters

probabilty (float) –

“Probability” is defined here as the forward-move

probability only. The backward-move probability is 1-probability.(Default = 0.25)

property probability

The probability of forward-move in an N-dimensional space

_walk(positions)

Run dynamics for the RandomWalk system for one step.

Parameters

positions (arraylike of shape (1, dimension)) – Current position of the walker.

Returns

new_positions – The positions of the walker after one dynamic step.

Return type

arraylike of shape (1, dimension)

post_cycle(**kwargs)

Perform post-cycle behavior. run_segment will be called for each walker so this allows you to perform changes of state on a per-cycle basis.

Parameters

kwargs (key-word arguments) – Key-value pairs to be interpreted by each runner implementation.

pre_cycle(**kwargs)

Perform pre-cycle behavior. run_segment will be called for each walker so this allows you to perform changes of state on a per-cycle basis.

Parameters

kwargs (key-word arguments) – Key-value pairs to be interpreted by each runner implementation.

run_segment(walker, segment_length, **kwargs)

Runs a random walk simulation for the given number of steps.

Parameters

walker (object implementing the Walker interface) – The walker for which dynamics will be propagated.

segment_lengthint

The numerical value that specifies how much dynamical steps are to be run.

Returns

new_walker – Walker after dynamics was run, only the state should be modified.

Return type

object implementing the Walker interface