piglot.objectives.response_objective.ScalarisationComposition

class ScalarisationComposition(objectives: List[ResponseSingleObjective], scalarisation: Scalarisation | None = None)

Bases: ResponseComposition

Composition for scalarisation of non-composite response objectives.

Methods

composition	Abstract method for computing the outer function of the composition
composition_torch	Compute the composition for all objectives.
get_latent_space	Compute the latent space representation of the given results.

composition(inner: ndarray, params: ndarray) → ndarray

Abstract method for computing the outer function of the composition

Parameters

innernp.ndarray

Return value from the inner function

paramsnp.ndarray

Parameters for the given responses

Returns

np.ndarray

Composition result

composition_torch(inner: Tensor, params: Tensor) → Tensor

Compute the composition for all objectives.

Parameters

innertorch.Tensor

Return value from the inner function.

paramstorch.Tensor

Paratemers for the given responses.

Returns

torch.Tensor

Composition results.

get_latent_space(params: ndarray, raw_responses: Dict[str, OutputResult]) → Tuple[ndarray, ndarray]

Compute the latent space representation of the given results.

Parameters

paramsnp.ndarray

Parameter values for these results.

raw_responsesDict[str, OutputResult]

Raw responses from the solver.

Returns

Tuple[np.ndarray, np.ndarray]

Latent space representation of the results: mean and covariance.