tempor.models.clairvoyance2.treatment_effects.synctwin module¶

class tempor.models.clairvoyance2.treatment_effects.synctwin.SyncTwinTensors(x_full, t_full, mask_full, batch_ind_full, y_full, y_control, y_mask_full)[source]¶

Bases: tuple

Create new instance of SyncTwinTensors(x_full, t_full, mask_full, batch_ind_full, y_full, y_control, y_mask_full)

x_full : Tensor¶

Alias for field number 0

t_full : Tensor¶

Alias for field number 1

mask_full : Tensor¶

Alias for field number 2

batch_ind_full : Tensor¶

Alias for field number 3

y_full : Tensor¶

Alias for field number 4

y_control : Tensor¶

Alias for field number 5

y_mask_full : Tensor¶

Alias for field number 6

class tempor.models.clairvoyance2.treatment_effects.synctwin.SyncTwinRegressor(params: dict[str, Any] | None = None)[source]¶

Bases: TreatmentEffectsModel, OrganizedTreatmentEffectsModuleMixin, OrganizedPredictorModuleMixin, OrganizedModule

requirements : Requirements = Requirements(dataset_requirements=DatasetRequirements(requires_static_covariates_present=False, requires_no_missing_data=True, static_covariates_value_type=<DataValueOpts.NUMERIC: 2>, temporal_covariates_value_type=<DataValueOpts.NUMERIC: 2>, temporal_targets_value_type=<DataValueOpts.NUMERIC: 2>, temporal_treatments_value_type=<DataValueOpts.ANY: 1>, event_covariates_value_type=<DataValueOpts.ANY: 1>, event_targets_value_type=<DataValueOpts.ANY: 1>, event_treatments_value_type=<DataValueOpts.NUMERIC_BINARY: 4>, requires_all_temporal_data_samples_aligned=False, requires_all_temporal_data_regular=False, requires_all_temporal_data_index_numeric=False, requires_all_temporal_containers_shares_index=True), prediction_requirements=PredictionRequirements(target_data_structure=<DataStructureOpts.TIME_SERIES: 1>, horizon_type=<HorizonOpts.TIME_INDEX: 2>, min_timesteps_target_when_fit=1, min_timesteps_target_when_predict=1), treatment_effects_requirements=TreatmentEffectsRequirements(treatment_data_structure=<DataStructureOpts.EVENT: 3>, min_timesteps_treatment_when_fit=1, min_timesteps_treatment_when_predict=1, min_timesteps_treatment_when_predict_counterfactual=1))¶

DEFAULT_PARAMS : dict[str, Any] | NamedTuple = _DefaultParams(hidden_size=20, tau=1.0, lambda_prognostic=1.0, lambda_reconstruction=1.0, batch_size=32, pretraining_iterations=5000, matching_iterations=20000, inference_iterations=20000, use_validation_set_in_training=True, treatment_status_is_treated=1)¶

expected_treatment_statuses = (0, 1)¶

property treated_indicator : int¶

property control_indicator : int¶

get_possible_prediction_horizon(sample_index: int, data: Dataset)[source]¶

get_possible_treatment_scenarios(sample_index: int, data: Dataset)[source]¶