Changeset c93e58 for src/FunctionApproximation/TrainingData.cpp

Timestamp:

Oct 2, 2016, 2:03:57 PM (9 years ago)

Author:

Frederik Heber <heber@…>

Branches:

Fix_FitPotential_needs_atomicnumbers

Children:

be20db

Parents:

6ffdf2

Message:

FIX: TrainingData used charges instead of atomic numbers to recognize fragments.

• this only works if we do not DoSampleValenceOnly. Otherwise charges are no longer directly associable with the atomic number of the nuclei.
• Extractors::gatherAllSymmetricDistanceArguments now needs atomicnumbers instead of charges.
• TESTFIX: needed to adapt ExtractorsUnitTest.

File:

• src/FunctionApproximation/TrainingData.cpp (modified) (1 diff)

src/FunctionApproximation/TrainingData.cpp

@@ -58 +58 @@
   for (HomologyContainer::const_iterator iter = range.first; iter != range.second; ++iter) {
     const Fragment &fragment = iter->second.fragment;
-    // create internal list of arguments
     FunctionModel::arguments_t all_args = Extractors::gatherAllSymmetricDistances(
         fragment.getPositions(),
-        fragment.getCharges(),
+        fragment.getAtomicNumbers(),
         DistanceVector.size()
         );