626_privacy/tensorflow_privacy
2
0
Fork 1
Code Issues Pull requests 1 Projects Releases Packages Wiki Activity Actions

Add slice names for custom slices

Browse source 
PiperOrigin-RevId: 544599507
This commit is contained in:
Vadym Doroshenko 2023-06-30 02:32:29 -07:00 committed by A. Unique TensorFlower
parent f953e834df
commit 93f5a5249c
2 changed files with 14 additions and 4 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

8
tensorflow_privacy/privacy/privacy_tests/membership_inference_attack/data_structures.py
View file

@ -20,7 +20,7 @@ import glob
import logging import logging
import os import os
import pickle import pickle
from typing import Any, Iterable, MutableSequence, Optional, Union, Sequence from typing import Any, Dict, Iterable, MutableSequence, Optional, Sequence, Union
import numpy as np import numpy as np
import pandas as pd import pandas as pd
@ -116,6 +116,12 @@ class SlicingSpec:
all_custom_train_indices: Optional[Sequence[np.ndarray]] = None all_custom_train_indices: Optional[Sequence[np.ndarray]] = None
all_custom_test_indices: Optional[Sequence[np.ndarray]] = None all_custom_test_indices: Optional[Sequence[np.ndarray]] = None
# Specifies names for custom slices. The names will be in the output in
# SingleAttackResult.slice_spec.value. If provided, the dictionary must
# contain names for all custom indices groups from all_custom_train_indices
# and all_custom_test_indices.
custom_slices_names: Optional[Dict[int, str]] = None
def __post_init__(self): def __post_init__(self):
if not self.all_custom_train_indices and not self.all_custom_test_indices: if not self.all_custom_train_indices and not self.all_custom_test_indices:
return return

10
tensorflow_privacy/privacy/privacy_tests/membership_inference_attack/dataset_slicing.py
View file

@ -264,9 +264,13 @@ def get_single_slice_specs(
f"Too many groups ({groups.size}) for slicing by custom indices. " f"Too many groups ({groups.size}) for slicing by custom indices. "
f"Should be no more than {_MAX_NUM_OF_SLICES}.") f"Should be no more than {_MAX_NUM_OF_SLICES}.")
for g in groups: for g in groups:
result.append( if slicing_spec.custom_slices_names is not None:
SingleSliceSpec(SlicingFeature.CUSTOM, if g not in slicing_spec.custom_slices_names:
(custom_train_indices, custom_test_indices, g))) raise ValueError(f"Custom slice={g} is not in custom_slices_names")
group_id = slicing_spec.custom_slices_names[g]
else:
group_id = (custom_train_indices, custom_test_indices, g)
result.append(SingleSliceSpec(SlicingFeature.CUSTOM, group_id))
return result return result