mlmc.sampler.Sampler

class mlmc.sampler.Sampler(sample_storage, sampling_pool, sim_factory, level_parameters, seed=1234)[source]

Manages sample scheduling, result collection, and persistent storage.

Coordinates the sampling pool, simulation factory, and sample storage: - schedules new samples according to target counts, - collects finished samples and writes them to storage, - handles failed samples and runtime (n_ops) bookkeeping.

__init__(sample_storage, sampling_pool, sim_factory, level_parameters, seed=1234)[source]

Initialize sampler and prepare per-level simulation objects.

Parameters:

  • sample_storage (SampleStorage) – store scheduled samples, results and result structure

  • sampling_pool (SamplingPool) – sampling pool responsible for executing simulations

  • sim_factory (Simulation) – factory that creates level Simulation instances and provides result_format()

  • level_parameters (List[List[float]]) – List of per-level parameters (e.g. simulation steps)

  • seed – global RNG seed used to seed NumPy’s RNG

Methods

__init__(sample_storage, sampling_pool, ...)

Initialize sampler and prepare per-level simulation objects.

ask_sampling_pool_for_samples([sleep, timeout])

Poll the sampling pool for finished simulations and store their results.

l_scheduled_samples()

Return the currently scheduled sample counts per level.

process_adding_samples(n_estimated[, sleep, ...])

Add newly estimated samples in batches, scheduling a fraction of the difference between current scheduled and newly estimated targets.

renew_failed_samples()

Reschedule previously failed samples.

sample_range(n0, nL)

Generate a geometric sequence of length L decreasing from n0 to nL.

schedule_samples([timeout, level_id, n_samples])

Schedule new simulation samples in the sampling pool and record them in storage.

set_initial_n_samples([n_samples])

Set initial target number of samples for each level.

set_level_target_n_samples(n_samples)

Update the per-level target sample counts to at least the provided values.

set_scheduled_and_wait(n_scheduled, ...[, ...])

Set scheduled sample targets and wait until a proportion of those samples finish.

Attributes

ADDING_SAMPLES_TIMEOUT

n_finished_samples

Retrieve numbers of finished samples for all levels.

n_levels

Return number of MLMC levels managed by this sampler.