Changeset 82e5fb for doc/userguide

Timestamp:

May 25, 2021, 11:04:50 PM (4 years ago)

Author:

Frederik Heber <frederik.heber@…>

Branches:

Candidate_v1.7.0, stable

Children:

91793c

Parents:

889067e

git-author:

Frederik Heber <frederik.heber@…> (05/25/21 21:55:59)

git-committer:

Frederik Heber <frederik.heber@…> (05/25/21 23:04:50)

Message:

Added option error-file to fit potential actions.

• DOCU: Added option to userguide.
• TESTS: Added option to FitCompoundPotential regression test.

File:

• doc/userguide/userguide.xml (modified) (2 diffs)

doc/userguide/userguide.xml

@@ -2430 +2430 @@
       --set-threshold 1e-3 \
       --training-file test.dat
+      --error-file error.dat
    </programlisting>
           <para>Now, all empirical potential functions are summed up into a
@@ -2437 +2438 @@
           potential between oxygen and hydrogen and another angular potential
           for the angle between hydrogen, oxygen, and hydrogen atom. Then, we would
-          fit a function consisting of the sum of the two potentials functions in order to approximate the energy of a single
-          water molecule (actually, it&apos;s the sum of three potentials. As mentioned before, a constant potential is always added to compensate non-bonding energies, i.e. not depending on interatomic distances). Here, the threshold criterion takes the place of the
-<emphasis role="bold"> take-best-of</emphasis> option. Here, the minimization is reiterated so often on random (but to some extent chosen from a sensible range) starting parameters until the final L2 error is below 1e-3. Also, all data used
-          for training, i.e. the tuples consisting of the fragments nuclei
-          coordinates and the associated energy value are written to the file
-          <filename>test.dat</filename>. This allows for graphical representation or other
-way of analysis, e.g. for a Morse potential between oxygen and hydrogen the bonding energy can be plotted as a one-dimensional function and compared to the &quot;point cloud&quot; of sample points from the fragment term of Born-Oppenheimer surface. It is this point cloud, i.e. the training data, that is written to the file
-          <filename>test.dat</filename>.</para>
+          fit a function consisting of the sum of the two potentials functions in 
+          order to approximate the energy of a single water molecule (actually, 
+          it&apos;s the sum of three potentials. As mentioned before, a constant 
+          potential is always added to compensate non-bonding energies, i.e. not 
+          depending on interatomic distances). Here, the threshold criterion takes 
+          the place of the<emphasis role="bold"> take-best-of</emphasis> option. 
+          Here, the minimization is reiterated so often on random (but to some 
+          extent chosen from a sensible range) starting parameters until the final 
+          L2 error is below 1e-3. Also, all data used for training, i.e. the tuples 
+          consisting of the fragments nuclei coordinates and the associated energy 
+          value are written to the file <filename>test.dat</filename>. This allows 
+          for graphical representation or other way of analysis, e.g. for a Morse 
+          potential between oxygen and hydrogen the bonding energy can be plotted 
+          as a one-dimensional function and compared to the &quot;point cloud&quot; 
+          of sample points from the fragment term of Born-Oppenheimer surface. 
+          It is this point cloud, i.e. the training data, that is written to the file
+          <filename>test.dat</filename>. Moreover, the error per fragment is written
+          to the file <filename>error.dat</filename></para>
           <para>Note that you can combine the two ways, i.e. start with a
           fit-potential call but give an empty potential file. The resulting