saealib.GenerationBasedStrategy#

class saealib.GenerationBasedStrategy(gen_ctrl)[source]#

Bases: OptimizationStrategy

Generation-based strategy.

Parameters:

gen_ctrl (int) – Number of surrogate-only generations executed inside each step() call before one generation of true objective evaluation.

Methods

__init__

build_graph

Lower the structured generation-based Pipeline to its graph.

build_pipeline

Describe the strategy as a structured Pipeline DSL.

contract

Return the strategy contract.

step

Run gen_ctrl surrogate-only generations, then one true-evaluation step.

Method Details

GenerationBasedStrategy.__init__(gen_ctrl)[source]#
Parameters:

gen_ctrl (int)

Return type:

None

GenerationBasedStrategy.build_graph(provider)[source]#

Lower the structured generation-based Pipeline to its graph.

Parameters:

provider (ComponentProvider)

Return type:

ComponentGraph

GenerationBasedStrategy.build_pipeline(provider)[source]#

Describe the strategy as a structured Pipeline DSL.

Parameters:

provider (ComponentProvider)

Return type:

Pipeline

GenerationBasedStrategy.contract()#

Return the strategy contract.

Return type:

ComponentContract

GenerationBasedStrategy.step(ctx, provider)[source]#

Run gen_ctrl surrogate-only generations, then one true-evaluation step.

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

Parameters:

  • ctx (OptimizationState) – Current optimization context.

  • provider (ComponentProvider) – Component provider.

Return type:

OptimizationState