selector.hp_point_selection

This module contains point selection functions.

Functions

distance_stats(smfeatures, distances)	Compute distance statistics.
get_relatives(suggested)	Get information of relations of suggested points by generator tag.
normalize_plus_cond_acc(sugg, s)	Normalize and account for conditionals.
pairwise_distances(sugg_i, sugg_j)	Compute pairwise distances.
select_point(scenario, suggested, max_evals, ...)	Generate features and run simulation.
simulation(suggested, features, max_evals, ...)	Run simulations of config selection.

selector.hp_point_selection.distance_stats(smfeatures, distances)

Compute distance statistics.

Parameters:

• suggested (list of selector.pool.Configuration) – List of suggested configurations.

• distances (list) – Distance values.

Returns:

New features for simulation.

Return type:

ndarray

selector.hp_point_selection.get_relatives(suggested)

Get information of relations of suggested points by generator tag.

Parameters:

suggested (list of selector.pool.Configuration) – List of suggested points.

Returns:

Nested array, indices of related points (by selector.pool.Generator).

Return type:

ndarray

selector.hp_point_selection.normalize_plus_cond_acc(sugg, s)

Normalize and account for conditionals.

Parameters:

• sugg (list of selector.pool.Configuration) – Suggested configurations.

• s (selector.scenario.Scenario) – AC scenario.

Returns:

Suggested configuration with normalized and adjusted values.

Return type:

list

selector.hp_point_selection.pairwise_distances(sugg_i, sugg_j)

Compute pairwise distances.

Parameters:

• sugg_i (list) – Configuration values for the first set.

• sugg_j (list) – Configuration values for the second set.

Returns:

Pairwise distances between the configurations.

Return type:

ndarray

selector.hp_point_selection.select_point(scenario, suggested, max_evals, npoints, pool, epoch, max_epoch, features, weights, seed)

Generate features and run simulation.

Parameters:

• s (selector.scenario.Scenario) – AC scenario.

• suggested (list) – List of configs/points to select from.

• max_eval (int) – Number of simulation runs per selected point.

• npoints (int) – Number of configs/points requested.

• pool (list) – List of configs/points to select from.

• epoch (int) – Current epoch.

• max_epoch (int) – Total number of epochs.

• features (ndarray (n_suggestions, n_features)) – Features computed for each suggested configuration.

• weights (ndarray (n_suggestions, n_features)) – Preset weights for the scoring function of the selection mechanism,

• seed (int) – Random seed.

Returns:

IDs of selected configs/points.

Return type:

list

selector.hp_point_selection.simulation(suggested, features, max_evals, selected_points, weights, npoints, distances, relatives)

Run simulations of config selection.

Parameters:

• suggested (list) – List of configs/points to select from.

• features (list) – Nested list, features of configs/points.

• max_eval (int) – Number of simulation runs per selected point.

• selected_points (list) – Indices of configurations selected so far in the simulations.

• weights (ndarray) – Weights for the scoring function.

• npoints (int) – Number of configurations to select

• distances (ndarray) – Distance features between the configuraions.

• relatives (ndarray) – Indices of relative configurations.

Returns:

How often configs/points were selected in the simulation.

Return type:

ndarray