Fix leaveOneOut algorithm for when averaging is false

testing/evaluateKinetics.py

@@ -89,13 +89,23 @@ def getKineticsLeaveOneOut(family, averaging=True):
     exactKinetics={}
     approxKinetics={}
 
+    rootTemplate = family.getRootTemplate()
+
     for entryKey in family.rules.entries.keys():
         template = family.retrieveTemplate(entryKey.split(';'))
-        exactKineticsData, exactKineticsEntry=family.rules.estimateKinetics(template)
-        if exactKineticsData.comment:
-            # This means it was an averaged node so skip
+        exactKineticsEntry=family.rules.getRule(template)  # fetch the best matched rule
+        if exactKineticsEntry.rank == 0:
+            # Skip rank zero entries, because they are replaced by averaged values
+            # during model generation
             continue
 
+        exactKineticsData = exactKineticsEntry.data
+        if exactKineticsData.comment:
+            if 'training' not in exactKineticsData.comment:
+                # Averaged nodes have kinetics data comments, so skip them unless 
+                # They were created from training reactions
+                continue
+
         exactKinetics[entryKey] = exactKineticsData
 
         if averaging:
@@ -111,7 +121,12 @@ def getKineticsLeaveOneOut(family, averaging=True):
             # This takes significanty less time to run, but is not a true validation of the data, 
             # it is just testing our kinetics selection algorithm
 
-            originalEntry = family.rules.entries(entryKey)
+            # Skip the top node in this scheme, because nothing can predict it if removed
+
+            if template == rootTemplate:
+                continue
+
+            originalEntry = family.rules.entries[entryKey]
             family.rules.entries.pop(entryKey)
             approxKinetics[entryKey], approxKineticsEntry=family.rules.estimateKinetics(template)
             # Re-add the data back into the original family