piglot.objectives.design.DesignSingleObjective

class DesignSingleObjective(name: str, prediction: List[str], quantity: Reduction, maximise: bool = False, weight: float = 1.0, bounds: Tuple[float, float] | None = None, flatten_utility: FlattenUtility | None = None, prediction_transform: ResponseTransformer | None = None)[source]

Single objective for design optimisation objectives.

Methods

`evaluate`	Evaluate objective value for the given results.
`evaluate_from_latent_space`	Evaluate the objective value(s) from the latent space representation.
`latent_space`	Compute latent space representation of the given results (for composite objectives).
`plot`	Plot the response for this objective.
`read`	Read the objective spec from the configuration dictionary.

evaluate(params: ndarray, raw_results: Dict[str, OutputResult]) → Tuple[float, float]

Evaluate objective value for the given results.

Parameters

evaluate_from_latent_space(params: Tensor, latent_responses: Tensor) → Tensor

Evaluate the objective value(s) from the latent space representation.

paramstorch.Tensor: Parameter values for these results.
latent_responsestorch.Tensor: Response(s) in the latent space representation.

latent_space(raw_results: Dict[str, OutputResult]) → Tuple[ndarray, ndarray]

Compute latent space representation of the given results (for composite objectives).

Tuple[np.ndarray, np.ndarray]: Mean and variance of the latent space representation.

plot(axis: Axes, raw_results: Dict[str, OutputResult]) → Dict[Line2D, str][source]

Plot the response for this objective.

Dict[Line2D, str]: Mapping of lines to response names (for dynamically updating plots).

classmethod read(name: str, config: Dict[str, Any], output_dir: str) → DesignSingleObjective[source]

Read the objective spec from the configuration dictionary.