Tidy remaining fraud detection review fixes

Author: holgerroth

research/fsi-fraud-detection/stats/client.py

@@ -75,8 +75,6 @@ def load_data(self, fl_ctx: FLContext) -> Dict[str, pd.DataFrame]:
             )
             for path in test_data_paths:
                 self.log_info(fl_ctx, f"  - {path}")
-
-            assert len(test_data_paths) == 4, "Expected 4 test files, got " + str(len(test_data_paths))
         else:
             # Single test file
             if not os.path.isfile(test_data_path_pattern):

research/fsi-fraud-detection/train/central_train.py

@@ -110,11 +110,6 @@ def main():
         print(f"Found {len(test_data_paths)} test files matching pattern: {test_data_path_pattern}")
         for path in test_data_paths:
             print(f"  - {path}")
-
-        # assert len(test_data_paths) == 25, "Expected 25 test files, got " + str(len(test_data_paths))
-        assert len(test_data_paths) == 20, "Expected 20 test files, got " + str(
-            len(test_data_paths)
-        )  # Datasets later than 3/2/2026
     else:
         # Single test file
         if not os.path.isfile(test_data_path_pattern):

research/fsi-fraud-detection/train/client.py

@@ -41,13 +41,6 @@
 from nvflare.app_opt.pt.fedproxloss import PTFedProxLoss
 from nvflare.client.tracking import MLflowWriter
 
-# install dependencies
-# os.system("python -m pip install opacus")
-# os.system("python -m pip install captum")
-# result = os.system("python -m pip install numpy==1.26.4")
-# print(f"Pip Install Result: {result}")
-
-
 PATH = "pt_model.weights.pth"
 
 
@@ -154,9 +147,6 @@ def main():
         print(f"Found {len(test_data_paths)} test files matching pattern: {test_data_path_pattern}")
         for path in test_data_paths:
             print(f"  - {path}")
-
-        # assert len(test_data_paths) == 5, "Expected 5 test files, got " + str(len(test_data_paths))
-        assert len(test_data_paths) == 4, "Expected 4 test files, got " + str(len(test_data_paths))  # new data 3/3/2026
     else:
         # Single test file
         if not os.path.isfile(test_data_path_pattern):
@@ -566,8 +556,10 @@ def main():
             print("[WARNING] Skip SHAP with DP")
         if shap_metrics:
             print(f"SHAP computation completed. Used {shap_metrics['shap_samples_used']} samples.")
-        else:
+        elif run_shap:
             print("SHAP computation failed. Skipping SHAP metrics.")
+        else:
+            print("SHAP computation skipped for this round.")
         metrics["shap_metrics"] = shap_metrics
 
         # (6) construct trained FL model (A dict of {parameter name: parameter weights} from the PyTorch model)