def __new__(cls, *args, **kwargs):
cls.name = "rf-robustness"
attacks = ['RBA_Approx_RF_100']
invert_ds_eps = {}
for k, v in ds_eps.items():
invert_ds_eps.setdefault(v, []).append(
auto_var.get_var_shown_name("dataset", k))
grid_params = []
for ds in tree_datasets:
v, k = [ds], ds_eps[auto_var.get_var_shown_name("dataset", ds)]
models = [
'random_forest_100_d5',
f'adv_rf_100_{k}_d5',
f'robust_rf_100_{k}_d5',
f'advPruning_rf_100_{k}_d5',
]
grid_params.append({
'model': models,
'ord': [ATTACK_NORM],
'dataset': v,
'attack': attacks,
'random_seed': random_seed,
})
cls.grid_params = grid_params
return RobustExperiments.__new__(cls, *args, **kwargs)
def __new__(cls, *args, **kwargs):
cls.name = "dt-robustness"
attacks = ['pgd']
invert_ds_eps = {}
for k, v in ds_eps.items():
invert_ds_eps.setdefault(v, []).append(
auto_var.get_var_shown_name("dataset", k))
grid_params = []
for ds in tree_datasets:
v, k = [ds], ds_eps[auto_var.get_var_shown_name("dataset", ds)]
models = [
'logistic_regression',
f'adv_logistic_regression_{k}',
f'advPruning_logistic_regression_{k}',
]
grid_params.append({
'model': models,
'ord': [ATTACK_NORM],
'dataset': v,
'attack': attacks,
'random_seed': random_seed,
})
cls.grid_params = grid_params
return RobustExperiments.__new__(cls, *args, **kwargs)
def __new__(cls, *args, **kwargs):
cls.name = "dt-robustness"
attacks = ['pgd']
grid_params = []
for ds in tree_datasets:
v, k = [ds], ds_eps[auto_var.get_var_shown_name("dataset", ds)]
models = ['mlp', f'adv_mlp_{k}', f'advPruning_mlp_{k}']
grid_params.append({
'model': models,
'ord': [ATTACK_NORM],
'dataset': v,
'attack': attacks,
'random_seed': random_seed,
})
cls.grid_params = grid_params
return RobustExperiments.__new__(cls, *args, **kwargs)
def __new__(cls, *args, **kwargs):
cls.name = "rf-robustness_l2"
attacks = ['RBA_Approx_KNN_k3_50']
grid_params = []
for ds in datasets:
v, k = [ds], ds_eps[auto_var.get_var_shown_name("dataset", ds)]
models = [
'knn3',
f'adv_nn_k3_{k}',
f'robustv2_nn_k3_{k}',
f'advPruning_nn_k3_{k}',
]
grid_params.append({
'model': models,
'ord': [ATTACK_NORM],
'dataset': v,
'attack': attacks,
'random_seed': random_seed,
})
cls.grid_params = grid_params
return RobustExperiments.__new__(cls, *args, **kwargs)