Changeset e1fe7e for src/FunctionApproximation/TrainingData.hpp

Timestamp:

Jun 27, 2014, 9:32:55 PM (11 years ago)

Author:

Frederik Heber <heber@…>

Branches:

Action_Thermostats, Add_AtomRandomPerturbation, Add_FitFragmentPartialChargesAction, Add_RotateAroundBondAction, Add_SelectAtomByNameAction, Added_ParseSaveFragmentResults, AddingActions_SaveParseParticleParameters, Adding_Graph_to_ChangeBondActions, Adding_MD_integration_tests, Adding_ParticleName_to_Atom, Adding_StructOpt_integration_tests, AtomFragments, Automaking_mpqc_open, AutomationFragmentation_failures, Candidate_v1.5.4, Candidate_v1.6.0, Candidate_v1.6.1, ChangeBugEmailaddress, ChangingTestPorts, ChemicalSpaceEvaluator, CombiningParticlePotentialParsing, Combining_Subpackages, Debian_Package_split, Debian_package_split_molecuildergui_only, Disabling_MemDebug, Docu_Python_wait, EmpiricalPotential_contain_HomologyGraph, EmpiricalPotential_contain_HomologyGraph_documentation, Enable_parallel_make_install, Enhance_userguide, Enhanced_StructuralOptimization, Enhanced_StructuralOptimization_continued, Example_ManyWaysToTranslateAtom, Exclude_Hydrogens_annealWithBondGraph, FitPartialCharges_GlobalError, Fix_BoundInBox_CenterInBox_MoleculeActions, Fix_ChargeSampling_PBC, Fix_ChronosMutex, Fix_FitPartialCharges, Fix_FitPotential_needs_atomicnumbers, Fix_ForceAnnealing, Fix_IndependentFragmentGrids, Fix_ParseParticles, Fix_ParseParticles_split_forward_backward_Actions, Fix_PopActions, Fix_QtFragmentList_sorted_selection, Fix_Restrictedkeyset_FragmentMolecule, Fix_StatusMsg, Fix_StepWorldTime_single_argument, Fix_Verbose_Codepatterns, Fix_fitting_potentials, Fixes, ForceAnnealing_goodresults, ForceAnnealing_oldresults, ForceAnnealing_tocheck, ForceAnnealing_with_BondGraph, ForceAnnealing_with_BondGraph_continued, ForceAnnealing_with_BondGraph_continued_betteresults, ForceAnnealing_with_BondGraph_contraction-expansion, FragmentAction_writes_AtomFragments, FragmentMolecule_checks_bonddegrees, GeometryObjects, Gui_Fixes, Gui_displays_atomic_force_velocity, ImplicitCharges, IndependentFragmentGrids, IndependentFragmentGrids_IndividualZeroInstances, IndependentFragmentGrids_IntegrationTest, IndependentFragmentGrids_Sole_NN_Calculation, JobMarket_RobustOnKillsSegFaults, JobMarket_StableWorkerPool, JobMarket_unresolvable_hostname_fix, MoreRobust_FragmentAutomation, ODR_violation_mpqc_open, PartialCharges_OrthogonalSummation, PdbParser_setsAtomName, PythonUI_with_named_parameters, QtGui_reactivate_TimeChanged_changes, Recreated_GuiChecks, Rewrite_FitPartialCharges, RotateToPrincipalAxisSystem_UndoRedo, SaturateAtoms_findBestMatching, SaturateAtoms_singleDegree, StoppableMakroAction, Subpackage_CodePatterns, Subpackage_JobMarket, Subpackage_LinearAlgebra, Subpackage_levmar, Subpackage_mpqc_open, Subpackage_vmg, Switchable_LogView, ThirdParty_MPQC_rebuilt_buildsystem, TrajectoryDependenant_MaxOrder, TremoloParser_IncreasedPrecision, TremoloParser_MultipleTimesteps, TremoloParser_setsAtomName, Ubuntu_1604_changes, stable

Children:

550f2a

Parents:

16227a

git-author:

Frederik Heber <heber@…> (02/27/14 20:15:41)

git-committer:

Frederik Heber <heber@…> (06/27/14 21:32:55)

Message:

FunctionModel now uses list_of_arguments to split sequence of subsets of distances.

• this fixes ambiguities with the set of distances: Imagine the distances within a water molecule as OH (A) and HH (B). We then may have a sequence of argument_t as AABAAB. And with the current implementation of CompoundPotential::splitUpArgumentsByModels() we would always choose the latter (and more complex) model. Hence, we make two calls to TriplePotential_Angle, instead of calls twice to PairPotential_Harmonic for A, one to PairPotential_Harmonic for B, and once to TriplePotential_Angle for AAB.
• now, we new list looks like A,A,B,AAB where each tuple of distances can be uniquely associated with a specific potential.
• changed signatures of EmpiricalPotential::operator(), ::derivative(), ::parameter_derivative(). This involved changing all of the current specific potentials and CompoundPotential.
• TrainingData must discern between the InputVector_t (just all distances) and the FilteredInputVector_t (tuples of subsets of distances).
• FunctionApproximation now has list_of_arguments_t as parameter to evaluate() and evaluate_derivative().
• DOCU: docu change in TrainingData.

File:

• src/FunctionApproximation/TrainingData.hpp (modified) (5 diffs)

src/FunctionApproximation/TrainingData.hpp

@@ -27 +27 @@
  * Fragment.
  *
- * In TrainingData::operator() we construct first all pair-wise distances as
- * list of all arguments. Then, these are filtered depending on the specific
- * FunctionModel's Filter and only these are handed to down to evaluate it.
+ * TrainingData::operator() takes the set of all possible pair-wise  distances
+ * (InputVector_t) and transforms it via the given filter into a list of subsets
+ * of distances (FilteredInputVector_t) that is feedable to the model.
  *
  */
@@ -41 +41 @@
   //!> Training tuple input vector pair
   typedef FunctionApproximation::inputs_t InputVector_t;
+  //!> Training tuple modified input vector pair
+  typedef FunctionApproximation::filtered_inputs_t FilteredInputVector_t;
   //!> Training tuple output vector pair
   typedef FunctionApproximation::outputs_t OutputVector_t;
@@ -47 +49 @@
   //!> Typedef for a map of each fragment with error.
   typedef std::multimap< double, size_t > L2ErrorConfigurationIndexMap_t;
-
 
 public:
@@ -75 +76 @@
    * \return const ref to training tuple of input vector
    */
-  const InputVector_t& getTrainingInputs() const {
+  const FilteredInputVector_t& getTrainingInputs() const {
     return ArgumentVector;
   }
@@ -142 +143 @@
   OutputVector_t EnergyVector;
   //!> list of all filtered arguments over all tuples
-  InputVector_t ArgumentVector;
+  FilteredInputVector_t ArgumentVector;
   //!> function to be used for training input data extraction from a fragment
   const FunctionModel::filter_t filter;