# Configuration file for the propensity sensitivity experiment
# Also compares the unbalancedness of the treatment assignment
# ==============================================================================

# Defaults
defaults:
  - _self_

  # Simulation settings
  - /simulator: ty_simulator

  # Model parametrization
  - /models@EconML_CausalForestDML: EconML_CausalForestDML
  - /models@EconML_DML: EconML_DML
  - /models@EconML_DMLOrthoForest: EconML_DMLOrthoForest
  - /models@EconML_DRLearner: EconML_DRLearner
  - /models@EconML_DROrthoForest: EconML_DROrthoForest
  - /models@EconML_ForestDRLearner: EconML_ForestDRLearner
  - /models@EconML_LinearDML: EconML_LinearDML
  - /models@EconML_LinearDRLearner: EconML_LinearDRLearner
  - /models@EconML_SparseLinearDML: EconML_SparseLinearDML
  - /models@EconML_SparseLinearDRLearner: EconML_SparseLinearDRLearner #EconML_SparseLinearDRLearner
  - /models@EconML_SLearner_Lasso: EconML_SLearner_Lasso
  - /models@EconML_TLearner_Lasso: EconML_TLearner_Lasso
  - /models@EconML_XLearner_Lasso: EconML_XLearner_Lasso

  - /models@Torch_SLearner: Torch_SLearner
  - /models@Torch_TLearner: Torch_TLearner
  - /models@Torch_XLearner: Torch_XLearner
  - /models@Torch_DRLearner: Torch_DRLearner
  - /models@Torch_RLearner: Torch_RLearner
  - /models@Torch_TARNet: Torch_TARNet
  - /models@Torch_DragonNet: Torch_DragonNet
  - /models@Torch_DragonNet_2: Torch_DragonNet_2
  - /models@Torch_DragonNet_4: Torch_DragonNet_4
  - /models@Torch_ULearner: Torch_ULearner
  - /models@Torch_RALearner: Torch_RALearner
  - /models@Torch_PWLearner: Torch_PWLearner
  - /models@Torch_FlexTENet: Torch_FlexTENet
  - /models@Torch_CRFNet_0_01: Torch_CRFNet_0_01
  - /models@Torch_CRFNet_0_001: Torch_CRFNet_0_001
  - /models@Torch_CRFNet_0_0001: Torch_CRFNet_0_0001
  - /models@Torch_ActionNet: Torch_ActionNet

  - /models@DiffPOLearner: DiffPOLearner


# EXPERIMENT AND DATA
# ==============================================================================
experiment_name: "propensity_scale_sensitivity"
# ==============================================================================

# EXPERIMENTAL KNOB
# ==============================================================================
# Cohort sizes to be tested
propensity_scales: [0, 0.25, 0.5, 1, 2, 4, 8, 16] #[0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6] #[0, 0.5, 1.0] #, 1.0]
unbalancedness_exps: [0] #, 0.2] # If too unbalanced, the model may give an error because of there being too few instances for a certain class
nonlinearity_scales: [0] #,0.4,0.8]
propensity_types: ["none_prog", "none_tre", "none_pred", "rct_none", "none_pred_overlap"] #, "none_prog_overlap"]
propensity_alpha: 1
n_samples: 100 #Irrelevant here, only for toy
# ==============================================================================