selector.pointselector

This module contains functions for selection of points.

Classes

HyperparameterizedSelector()

Hyperparameterized selection of generated points.

PointSelector([features])

Generic point selector class.

RandomSelector()

Random point selector class.
class selector.pointselector.HyperparameterizedSelector[source]

Bases: PointSelector

Hyperparameterized selection of generated points.

Note

Based on: Carlos Ansótegui, Meinolf Sellmann, Tapan Shah, Kevin Tierney, Learning to Optimize Black-Box Functions With Extreme Limits on the Number of Function Evaluations, 2021, International Conference on Learning and Intelligent Optimization, 7-24

select_points(scenario, pool, number_of_points, epoch, max_epoch, features, weights, results, max_evals=100, seed=False)[source]

Select a subset of configurations from the pool based on a scoring function.

Parameters:

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

  • pool (dict) – Pool of configurations to select from.

  • number_of_points (int) – Number of points to select from the pool.

  • epcoch (int) – Iteration identifier which stores the selection for later reference.

  • max_epoch (int) – Maximum number of iterations for the AC process (meaningless if termination criterion is total_runtime).

  • features (ndayrray) – Configuration features computed for each configuration in the pool.

  • weights (ndarray) – Pre-computed/ set weights for the scoring fuction.

  • results (dict) – Results for configuration /instance pairs.

  • max_evals (int) – Number of simulations per selected point.

  • seed (int) – Random seed.

Returns:

IDs of configurations from the pool that are selected.

Return type:

list

class selector.pointselector.PointSelector(features=None)[source]

Bases: object

Generic point selector class.

Parameters:

features (list of dict) – Problem instance features.

select_points()[source]

Generic point selector method.

class selector.pointselector.RandomSelector[source]

Bases: PointSelector

Random point selector class.

select_points(pool, number_of_points, iteration, seed=False)[source]

Randomly select a subset of configurations from the pool to run.

Parameters:

  • pool (dict) – Pool of configurations to select from.

  • number_of_points (int) – Number of points to select from the pool.

  • iteration (int) – Iteration identifier which stores the selection for later reference.

  • seed (int) – Random seed.

Returns:

IDs of configurations from the pool that are selected.

Return type:

list