Skip to content
voiage

voiage.methods.federated_privacy_preserving.value_of_federated_privacy_preserving

value_of_federated_privacy_preserving

Section titled “value_of_federated_privacy_preserving”
value_of_federated_privacy_preserving([positional or keyword] site_summaries: np.ndarray | list[list[float]] = None, [keyword-only] site_weights: np.ndarray | list[float] | None = None, [keyword-only] privacy_budgets: np.ndarray | list[float] = None, [keyword-only] prior_strategy_values: np.ndarray | list[float] = None, [keyword-only] strategy_names: list[str] | None = None, [keyword-only] synthetic_site_summaries: np.ndarray | list[list[float]] | None = None, [keyword-only] noise_scale: float = 0.0, [keyword-only] information_cost: float = 0.0, [keyword-only] individual_data_access: str = 'blocked', [keyword-only] seed: int = 0) -> FederatedPrivacyPreservingResult

Calculate VOI from site-local summaries without centralizing records.

Summaries are weighted across sites after optional deterministic Laplace noise calibrated by each site’s privacy budget. If individual-level access is blocked and synthetic summaries are supplied, the synthetic matrix is used and recorded in diagnostics. The method is fixture-backed pending external privacy review, binding parity, and licensed multi-site data.

Parameters:

  • site_summaries np.ndarray | list[list[float]]
  • site_weights np.ndarray | list[float] | None (default: None)
  • privacy_budgets np.ndarray | list[float]
  • prior_strategy_values np.ndarray | list[float]
  • strategy_names list[str] | None (default: None)
  • synthetic_site_summaries np.ndarray | list[list[float]] | None (default: None)
  • noise_scale float (default: 0.0)
  • information_cost float (default: 0.0)
  • individual_data_access str (default: 'blocked')
  • seed int (default: 0)

Returns: FederatedPrivacyPreservingResult