Add option to perturbation test script to force the perturbation of s…

…pecific attributes first

didactic/apps/perturbation_test.py

@@ -95,6 +95,7 @@ def run_perturbation_test(
     patients: Patients,
     target: TabularAttribute,
     perturbation_mode: Literal["negative", "positive"] = "negative",
+    manual_perturbations: Sequence[TabularAttribute] = None,
     mask_tag: str = CardinalTag.mask,
     progress_bar: bool = False,
     save_dir: Path = None,
@@ -107,6 +108,7 @@ def run_perturbation_test(
         target: Target attribute w.r.t. which to compute the model's AUROC score.
         perturbation_mode: Type of perturbation test to perform. If ``negative``, the test will remove the least
             relevant attributes first. If ``positive``, the test will remove the most relevant attributes first.
+        manual_perturbations: Attributes to manually perturb before automatically selecting following attributes.
         mask_tag: Tag of the segmentation mask for which to extract the time-series attributes.
         progress_bar: If ``True``, enables progress bars detailing the progress of how many attributes are left to
             perturb.
@@ -115,6 +117,9 @@ def run_perturbation_test(
     Returns:
         A series containing the AUROC score of the model for each further attribute removed.
     """
+    if manual_perturbations is None:
+        manual_perturbations = []
+
     tab_attrs, time_series_attrs = model.hparams.tabular_attrs, model.hparams.time_series_attrs
     n_attrs = len(tab_attrs) + (len(time_series_attrs) * len(model.hparams.views))
 
@@ -190,8 +195,13 @@ def run_perturbation_test(
             remaining_attrs = attrs_relevancy[~attrs_relevancy.index.isin(attrs_to_remove)]
             if remaining_attrs.empty:
                 break
+
+            if manual_perturbations:
+                next_attr = manual_perturbations.pop(0)
             else:
-                attrs_to_remove.append(remaining_attrs.index[0])
+                next_attr = remaining_attrs.index[0]
+
+            attrs_to_remove.append(next_attr)
 
     return pd.Series(attrs_perturbation_scores, name="AUROC")
 
@@ -229,6 +239,12 @@ def main():
         help="Type of perturbation test to perform. 'negative' removes the least relevant attributes first, "
         "'positive' removes the most relevant attributes first",
     )
+    parser.add_argument(
+        "--manual_perturbations",
+        type=TabularAttribute,
+        nargs="*",
+        help="Attributes to manually perturb before automatically selecting following attributes",
+    )
     parser.add_argument(
         "--mask_tag",
         type=str,
@@ -239,10 +255,11 @@ def main():
     args = parser.parse_args()
     kwargs = vars(args)
 
-    encoder_ckpt, relevancy_target, perturbation_mode, mask_tag, output_dir = (
+    encoder_ckpt, relevancy_target, perturbation_mode, manual_perturbations, mask_tag, output_dir = (
         kwargs.pop("pretrained_encoder"),
         kwargs.pop("relevancy_target"),
         kwargs.pop("perturbation_mode"),
+        kwargs.pop("manual_perturbations"),
         kwargs.pop("mask_tag"),
         kwargs.pop("output_dir"),
     )
@@ -256,6 +273,7 @@ def main():
         patients,
         relevancy_target,
         perturbation_mode=perturbation_mode,
+        manual_perturbations=manual_perturbations,
         mask_tag=mask_tag,
         progress_bar=True,
         save_dir=output_dir,