Address final CWRU fixes for Issue #8

Author: TasfinMahmud

cross_domain_validation.py cross_condition_robustness.pycross_domain_validation.py renamed to cross_condition_robustness.py

@@ -23,7 +23,7 @@ def add_awgn(signals, snr_db):
 
 def main():
     print('=' * 80)
-    print('CNSD CROSS-DOMAIN VALIDATION (AWGN NOISE INJECTION)')
+    print('CNSD CROSS-CONDITION ROBUSTNESS (AWGN NOISE INJECTION)')
     print('=' * 80)
 
     # 1. Load baseline data
@@ -47,7 +47,7 @@ def main():
     model.fit(train_data, epochs=30)
     print('Training complete.\n')
 
-    # 3. Evaluate on noisy cross-domains
+    # 3. Evaluate on noisy cross-conditions
     snr_levels = [None, 0, -5, -10]
 
     for snr in snr_levels:

threshold_sweep.py

@@ -0,0 +1,43 @@
+"""
+threshold_sweep.py - sweep the envelope-prominence threshold on the CWRU dataset.
+"""
+import numpy as np
+from cnsd import Dataset
+from cnsd.diagnosis.system import CNSD
+
+# Reuse CWRU loader and configs from validate_run
+from validate_run import load_cwru, CWRU, TAXONOMY
+
+def main():
+    print('=' * 50)
+    print('CWRU THRESHOLD SWEEP')
+    print('=' * 50)
+
+    X, y, cond = load_cwru()
+    X = np.asarray(X, np.float32); y = np.asarray(y); cond = np.asarray(cond)
+    
+    test_mask = cond == 3
+    train_mask = cond < 3
+    
+    train_data = Dataset.from_arrays(X[train_mask], y[train_mask], cond[train_mask], fs=12000, physics=CWRU, taxonomy=TAXONOMY, name='CWRU_Train')
+    test_data = Dataset.from_arrays(X[test_mask], y[test_mask], cond[test_mask], fs=12000, physics=CWRU, taxonomy=TAXONOMY, name='CWRU_Test')
+    
+    print('Training baseline CNSD model (this happens once)...')
+    model = CNSD()
+    model.fit(train_data, epochs=30)
+    
+    print('\nStarting Threshold Sweep on Test Data:')
+    thresholds = [3.0, 2.5, 2.0, 1.5]
+    
+    for t in thresholds:
+        print('-' * 40)
+        print(f'Testing Threshold: {t}')
+        model.symbolic.tau = t  # Hot-swap the threshold!
+        report = model.diagnose(test_data)
+        vr = report.verification_rate()
+        print(f"  CONFIRMED   : {vr.get('CONFIRMED', 0.0):.1%}")
+        print(f"  CONFLICT    : {vr.get('CONFLICT', 0.0):.1%}")
+        print(f"  INCONCLUSIVE: {vr.get('INCONCLUSIVE', 0.0):.1%}")
+
+if __name__ == '__main__':
+    main()

validate_run.py

@@ -74,10 +74,11 @@ def read_mat(path, label, load):
         for size, label in size_map.items():
             dir_path = os.path.join(fault_dir, ftype, size)
             if not os.path.exists(dir_path): continue
-            for f in os.listdir(dir_path):
-                if f.endswith('.mat'):
-                    load = int(f.split('_')[-1].split('.')[0])
-                    read_mat(os.path.join(dir_path, f), label, load)
+            for root, dirs, files in os.walk(dir_path):
+                for f in files:
+                    if f.endswith('.mat'):
+                        load = int(f.split('_')[-1].split('.')[0])
+                        read_mat(os.path.join(root, f), label, load)
 
     return np.array(X, dtype=np.float32), np.array(y), np.array(cond)
 
@@ -174,9 +175,19 @@ def main():
               'Run: pip install dowhy')
 
     # 7. example auditable diagnoses
-    print('\n[examples] auditable root-cause diagnoses:')
-    for r in report.records[:5]:
-        print(f'    [{r["status"]}] {r["root_cause"]["statement"]}')
+    print('\n[examples] auditable root-cause diagnoses (CONFIRMED faults only):')
+    seen_classes = set()
+    examples = []
+    for idx, r in enumerate(report.records):
+        y_true = test_data.y[idx]
+        if r["physics_verdict"] == "CONFIRMED" and y_true > 0:
+            if y_true not in seen_classes:
+                examples.append(f'    [Class {y_true}] [{r["status"]}] {r["root_cause"]["statement"]}')
+                seen_classes.add(y_true)
+            if len(examples) >= 5:
+                break
+    for ex in examples:
+        print(ex)
 
     print('\n' + '=' * 68)
     print('VALIDATION COMPLETE - if all sections printed, the pipeline runs '