Millepede: add convergence as parameter

src/algorithms/Millepede/Millepede.cpp

@@ -23,6 +23,8 @@ Millepede::Millepede(Configuration config, std::vector<Detector*> detectors)
     m_rescut_init = m_config.get<double>("residual_cut_init", 0.6);
     m_nstdev = m_config.get<double>("NStdDev", 0);
 
+    m_convergence = m_config.get<double>("convergence", 0.00001);
+
     // Use default values for the sigmas, unless specified explicitly.
     m_sigmas = m_config.getArray<double>("sigmas", {0.05, 0.05, 0.5, 0.005, 0.005, 0.005});
 }
@@ -153,7 +155,7 @@ void Millepede::finalise() {
                 cluster->setClusterCentre(pGlobal.x(), pGlobal.y(), pGlobal.z());
             }
         }
-        if(converg < 0.00001)
+        if(converg < m_convergence)
             break;
     }
 }

src/algorithms/Millepede/Millepede.h

@@ -144,6 +144,8 @@ namespace corryvreckan {
         double m_cfactref;
         // Number of standard deviations for chisquare / ndof cut.
         int m_nstdev;
+        // Value of convergence to interrupting iterations
+        double m_convergence;
         /// Number of "full" iterations (with geometry updates).
         unsigned int m_nIterations;
         /// Sigmas for each degree of freedom

src/algorithms/Millepede/README.md

@@ -7,6 +7,7 @@ This implementation of the Millepede algorithm has been taken from the [Kepler f
 
 The Millepede algorthm allows a simultaneous fit of both the tracks and the alignment constants.
 
+The algorithms stops if the convergence, i.e. the absolute sum of all corrections over the total number of parameters, is smaller than the configured value.
 
 #### Parameters
 * `number_of_tracks`: Number of tracks used in the alignment method chosen. Default value is `20000`.
@@ -16,6 +17,7 @@ The Millepede algorthm allows a simultaneous fit of both the tracks and the alig
 * `residual_cut_init`: Initial residual cut for outlier rejection in the first iteration. This value is applied for the first iteration and replaced by `residual_cut` thereafter. Default value is `0.6mm`.
 * `nstddev`: Cut to reject track candidates based on their Chi2/ndof value. Default value is `0`, i.e. the feature is disabled.
 * `sigmas`: Uncertainties for each of the alignment parameters. Defaults to `0.05, 0.05, 0.5, 0.005, 0.005, 0.005`.
+* `convergence`: Convergence value at which the algorithm stops iterating. Default value is `10e-5`.
 
 #### Plots produced
 No plots are produced.