Changeset 322d58 for doc/userguide/userguide.xml

Timestamp:

Apr 4, 2018, 4:59:48 PM (7 years ago)

Author:

Frederik Heber <frederik.heber@…>

Branches:

AutomationFragmentation_failures, Candidate_v1.6.1, ChemicalSpaceEvaluator, Enhanced_StructuralOptimization, Enhanced_StructuralOptimization_continued, Exclude_Hydrogens_annealWithBondGraph, ForceAnnealing_with_BondGraph, ForceAnnealing_with_BondGraph_contraction-expansion, Gui_displays_atomic_force_velocity, JobMarket_RobustOnKillsSegFaults, JobMarket_StableWorkerPool, PythonUI_with_named_parameters, StoppableMakroAction, TremoloParser_IncreasedPrecision

Children:

bca6b6

Parents:

1e49e6

git-author:

Frederik Heber <heber@…> (04/06/17 05:09:37)

git-committer:

Frederik Heber <frederik.heber@…> (04/04/18 16:59:48)

Message:

ForceAnnealing now uses step width according to Barzilai-Borwein method.

• this is not as good as what MPQC does but as long as Actions don't have an internal state, we cannot do anything better, e.g. a real line-search.
• removed paramater delta from optimize-structure.
• DOCU: Updated entry in user guide and noted Barzilai-Borwein method.
• TESTS: removed deltat from force annealing regression tests.
• TESTS: switched off StructureOptimization integration tests as poolworkers crash at the moment, making the tests hang indefinitely.

File:

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

doc/userguide/userguide.xml

@@ -2431 +2431 @@
     combines the same Actions as <emphasis role="bold">molecular-dynamics</emphasis> does. However, it
     uses the <emphasis role="bold">force-annealing</emphasis> action instead of <emphasis role="bold">verlet-integration</emphasis>.</para>
+          <remark>Because it is a MacroAction we cannot use any more elaborate 
+          line search method during the optimization. In essence, we minimize the
+          energy function that depends on the nuclei coordinates. The function is
+          non-linear and non-convex. The best choice, given that we have gradient
+          information, would be the Conjugate Gradient method (such as Fletcher-
+          Reeves which work well also for non-linear functions). However, these
+          methods perform a line search along the gradient direction which we 
+          cannot as our Actions do not contain any internal information (apart
+          from the state for Undo/Redo). Therefore, we are restricted to the
+          method of Barzilai-Borwein where no line-search is needed but that also
+          works well for high-dimensional minimization problems.</remark>
           <para>The command below performs a structure optimization of the 
     currently selected atoms (may also be a subset) for up to 100 time
@@ -2450 +2461 @@
     Otherwise only two time steps would be created: the initial and
     the final one containing the optimized structure.</para>
+        <para>Because of the use of Barzilai-Borwein for computing the 
+        stepwidth  we do not need any 'deltat' parameters. If the computed
+        step width is zero, we use a default step width of 1.</para>
         </section>
         <section xml:id="dynamics.step-world-time">