saealib.IndividualBasedStrategy#

class saealib.IndividualBasedStrategy(evaluation_ratio=0.1)[ソース]#

ベースクラス: OptimizationStrategy

Individual-based strategy.

パラメータ:

evaluation_ratio (float)

Methods

__init__

Initialize IndividualBasedStrategy.

build_graph

Build the canonical individual-based strategy graph.

build_pipeline

Return the strategy's high-level Pipeline description.

contract

Return the strategy contract.

step

Score all offspring with the surrogate, then true-evaluate the top fraction.

Method Details

IndividualBasedStrategy.__init__(evaluation_ratio=0.1)[ソース]#

Initialize IndividualBasedStrategy.

パラメータ:

evaluation_ratio (float) -- Ratio of offspring selected for true objective evaluation. The top evaluation_ratio fraction of offspring are evaluated with the real objective function; the rest are discarded after surrogate scoring. The number of neighbors for local surrogate fitting is now configured on the SurrogateManager (e.g. LocalSurrogateManager).

IndividualBasedStrategy.build_graph(provider)[ソース]#

Build the canonical individual-based strategy graph.

パラメータ:

provider (ComponentProvider)

戻り値の型:

ComponentGraph

IndividualBasedStrategy.build_pipeline(provider)#

Return the strategy's high-level Pipeline description.

パラメータ:

provider (ComponentProvider)

戻り値の型:

Pipeline

IndividualBasedStrategy.contract()#

Return the strategy contract.

戻り値の型:

ComponentContract

IndividualBasedStrategy.step(ctx, provider)[ソース]#

Score all offspring with the surrogate, then true-evaluate the top fraction.

Rebuilds the pipeline each call so component/parameter changes take effect immediately.

パラメータ:
  • ctx (OptimizationState) -- Current optimization context.

  • provider (ComponentProvider) -- Component provider.

戻り値の型:

OptimizationState