EqxLearnUncertainty

class EqxLearnUncertainty(initial_models=10, batch_size=1, num_grid_per_dim=1024, grid_sampler=LatinHypercube(), surrogate=None, fit_kwargs=<factory>)

Bases: FieldSampler

Adaptive sampler targeting regions of high surrogate uncertainty.

Parameters:

• initial_models (int)

• batch_size (int)

• num_grid_per_dim (int)

• grid_sampler (AbstractOneShotSampler)

• surrogate (Any)

• fit_kwargs (dict)

surrogate: Any = None

fit_kwargs: dict

train_field(params, features, key)

Parameters:

• params (Array)

• features (Array)

• key (Array)

Return type:

Any

evaluate_field(field, theta, key)

Parameters:

• field (Any)

• theta (Array)

• key (Array)

Return type:

float

batch_size: int = 1

grid_sampler: AbstractOneShotSampler = LatinHypercube()

init(eval_fn, d, key, options)

Initialize the sampler state and perform the first evaluations.

Return type:

SamplerState

initial_models: int = 10

num_grid_per_dim: int = 1024

step(eval_fn, d, state, key, options)

Perform one iteration of the sampling algorithm.

Return type:

SamplerState

terminate(state, target_N)

Determine whether the sampling loop should stop.

Parameters:

• state (SamplerState)

• target_N (int)

Return type:

bool