variance_min > 0 => indiscernibles are used, otherwise max_degree is …

…very large

phylogeny/data/Salmonella-var_min.deformation_outliers

@@ -0,0 +1,22 @@
+SAMN02844477
+SAMN03475395
+SAMN05605405
+SAMN02844313
+SAMN03002022
+SAMN03478846
+SAMN04485948
+SAMN04535377
+SAMN02368790
+SAMN04308849
+SAMN03474009
+SAMN02699690
+SAMN05201703
+SAMN02640780
+SAMN02843948
+SAMN03838256.r2
+SAMN04102293.r2
+SAMN04363882
+SAMN03892264
+SAMN03154403
+SAMN02645767
+SAMN04507552

phylogeny/data/Salmonella-var_min.distTree

@@ -10,62 +10,66 @@ Neighbor joining ...
 
 INPUT:
 # Leaves: 291
-# Discernible leaves: 291
+# Discernible leaves: 285
 # Nodes: 580
 # Dissimilarities: 42195 (100.00 %)
-Dissimilarities factor: 25.56
-Ave. dissimilarity = 0.191845
-Abs. criterion = 1.8111e+01  Rel. criterion = 2.951 %
+Dissimilarities factor: 26.19
+Ave. dissimilarity = 0.197058
+Abs. criterion = 1.8217e+01  Rel. criterion = 2.959 %
 
 Optimizing arc lengths for the whole tree ...
 Optimizing arc lengths at each arc ...
 # Nodes deleted = 17
 Optimizing arc lengths at each node ...
-# Nodes deleted = 4
-Abs. criterion = 7.2358e+00  Rel. criterion = 1.865 %
+# Nodes deleted = 2
+Abs. criterion = 7.2835e+00  Rel. criterion = 1.871 %
 
 Optimizing topology: subgraphs ...
 # Hybrids: 0
-Abs. criterion = 6.5163e+00  Rel. criterion = 1.763 %
+Abs. criterion = 6.5364e+00  Rel. criterion = 1.766 %
 
 # Hybrids: 0
-Abs. criterion = 6.3905e+00  Rel. criterion = 1.746 %
+Abs. criterion = 6.5128e+00  Rel. criterion = 1.763 %
 
 # Hybrids: 0
-Abs. criterion = 6.2287e+00  Rel. criterion = 1.724 %
+Abs. criterion = 6.5049e+00  Rel. criterion = 1.762 %
 
 # Hybrids: 0
-Abs. criterion = 6.2112e+00  Rel. criterion = 1.722 %
+Abs. criterion = 6.5035e+00  Rel. criterion = 1.761 %
 
 # Hybrids: 0
-Abs. criterion = 6.2088e+00  Rel. criterion = 1.721 %
+Abs. criterion = 6.5029e+00  Rel. criterion = 1.761 %
 
 # Hybrids: 0
-Abs. criterion = 6.2084e+00  Rel. criterion = 1.721 %
+Abs. criterion = 6.5025e+00  Rel. criterion = 1.761 %
 
 # Hybrids: 0
-Abs. criterion = 6.2082e+00  Rel. criterion = 1.721 %
+Abs. criterion = 6.5023e+00  Rel. criterion = 1.761 %
 
 # Hybrids: 0
-Abs. criterion = 6.2081e+00  Rel. criterion = 1.721 %
+Abs. criterion = 6.5021e+00  Rel. criterion = 1.761 %
 
 # Hybrids: 0
-Abs. criterion = 6.2081e+00  Rel. criterion = 1.721 %
+Abs. criterion = 6.5020e+00  Rel. criterion = 1.761 %
 
-# Iterations of subgraph optimization: 9
-Abs. criterion = 6.2081e+00  Rel. criterion = 1.721 %
+# Hybrids: 0
+Abs. criterion = 6.5020e+00  Rel. criterion = 1.761 %
+
+# Iterations of subgraph optimization: 10
+Abs. criterion = 6.5019e+00  Rel. criterion = 1.761 %
 
 Ave. radius: 0.456
 
 # Criterion outliers: 4
-Abs. criterion = 4.8597e+00  Rel. criterion = 1.542 %
+Abs. criterion = 4.8721e+00  Rel. criterion = 1.543 %
 
-# Deformation outliers: 0
+# Deformation outliers: 22
+Abs. criterion = 3.6261e+00  Rel. criterion = 1.447 %
 
 OUTPUT:
-Abs. criterion = 4.8597e+00  Rel. criterion = 1.542 %
+Abs. criterion = 3.6261e+00  Rel. criterion = 1.447 %
 
 Relative epsilon2_0 = 0.00 %
 Mean residual = 0.00
-Correlation between residual^2 and dissimilarity = 0.09
+Correlation between residual^2 and dissimilarity = 0.12
 

phylogeny/data/tree4.stat

@@ -5,7 +5,7 @@
 Tree length = 15.000
 Min. discernible leaf length = 1.000000e+00
 Ave. arc length = 2.500
-Interior height = 1
+Height ignoring indiscernibles = 1
 Bifurcating interior branching = 2.000
 Max. degree = 3
 # Frequent children interior nodes = 3

phylogeny/distTree.cpp

@@ -610,7 +610,7 @@ void DTNode::qc () const
     QC_ASSERT ((bool) getParent () == ! isNan (len));
     if (! childrenDiscernible ())
     {
-      QC_ASSERT (! DistTree_sp::variance_min);
+    //QC_ASSERT (! DistTree_sp::variance_min);
       QC_ASSERT (! inDiscernible ());
       for (const DiGraph::Arc* arc : arcs [false])
         QC_ASSERT (static_cast <const DTNode*> (arc->node [false]) -> inDiscernible ());
@@ -665,7 +665,7 @@ const Leaf* DTNode::inDiscernible () const
       return nullptr;
     else
     {
-      ASSERT (! DistTree_sp::variance_min);
+    //ASSERT (! DistTree_sp::variance_min);
       return g;
     }
   else
@@ -1313,7 +1313,7 @@ void Leaf::qc () const
 
   QC_ASSERT (! name. empty());
   QC_ASSERT (! isLeft (name, "0x"));
-  QC_IMPLY (DistTree_sp::variance_min, discernible);
+//QC_IMPLY (DistTree_sp::variance_min, discernible);
   if (! isNan (len) && ! discernible && len)
   {
     cout << getName () << " " << len << endl;
@@ -1375,7 +1375,7 @@ void Leaf::collapse (Leaf* other)
 {
   ASSERT (! getDistTree (). subDepth);
   ASSERT (this != other);
-  ASSERT (! DistTree_sp::variance_min);
+//ASSERT (! DistTree_sp::variance_min);
 //ASSERT (len == 0.0);
 //ASSERT (discernible);
 
@@ -1608,8 +1608,8 @@ void Subgraph::removeIndiscernibles ()
 {
   ASSERT (! area. empty ());
 
-  if (DistTree_sp::variance_min)
-    return;
+//if (DistTree_sp::variance_min)
+  //return;
 
   for (Iter<VectorPtr<Tree::TreeNode>> iter (area); iter. next (); )
     if (static_cast <const DTNode*> (*iter) -> inDiscernible ())
@@ -2888,8 +2888,8 @@ struct DissimLine
         return;
       if (! leaf2)
         return;
-      if (   ! DistTree_sp::variance_min 
-          && ! dissim 
+      if (   /*! DistTree_sp::variance_min 
+          &&*/ ! dissim 
           && ! leaf1->getCollapsed (leaf2)  // Only for new Leaf's
          )  
         leaf1->collapse (leaf2);
@@ -3882,7 +3882,7 @@ bool DistTree::loadLines (const StringVector &lines,
   {
     ASSERT (parent);
     auto leaf = new Leaf (*this, parent, len, idS);
-    leaf->discernible = DistTree_sp::variance_min || ! indiscernible;
+    leaf->discernible = /*DistTree_sp::variance_min ||*/ ! indiscernible;
     leaf->normCriterion = normCriterion;
     dtNode = leaf;
   }
@@ -4191,8 +4191,8 @@ LeafCluster DistTree::getIndiscernibles ()
   ASSERT (! subDepth);
   ASSERT (optimizable ());
 
-  if (DistTree_sp::variance_min)
-    return LeafCluster ();
+//if (DistTree_sp::variance_min)
+  //return LeafCluster ();
 
   // Leaf::DisjointCluster
   for (DiGraph::Node* node : nodes)
@@ -4245,7 +4245,7 @@ size_t DistTree::leafCluster2discernibles (const LeafCluster &leafCluster)
     if (clusterNodes. size () == 1)
       continue;
 
-    ASSERT (! DistTree_sp::variance_min);
+  //ASSERT (! DistTree_sp::variance_min);
     const Leaf* first = clusterNodes [0];
     ASSERT (first);
     Steiner* parent = var_cast (static_cast <const DTNode*> (first->getParent ()) -> asSteiner ());
@@ -4321,8 +4321,8 @@ size_t DistTree::setDiscernibles_ds ()
     }
   }
 
-  if (DistTree_sp::variance_min)
-    return 0;
+//if (DistTree_sp::variance_min)
+  //return 0;
 
   FFOR (size_t, row, dissimDs->objs. size ())
     if (const Leaf* leaf1 = findPtr (name2leaf, dissimDs->objs [row] -> name))
@@ -4359,8 +4359,8 @@ size_t DistTree::setDiscernibles ()
       leaf->discernible = true;
   }  
 
-  if (DistTree_sp::variance_min)
-    return 0;
+//if (DistTree_sp::variance_min)
+  //return 0;
 
   const LeafCluster leafCluster (getIndiscernibles ());
   const size_t n = leafCluster2discernibles (leafCluster);
@@ -4873,7 +4873,7 @@ bool DistTree::addDissim (Leaf* leaf1,
   ASSERT (leaf1);
   ASSERT (leaf2);
   IMPLY (! multFixed, mult == 1.0);  // mult will be set later
-  IMPLY (! DistTree_sp::variance_min && ! target && dissimTypes. empty (), leaf1->getCollapsed (leaf2));  
+  IMPLY (/*! DistTree_sp::variance_min &&*/ ! target && dissimTypes. empty (), leaf1->getCollapsed (leaf2));  
 
 
   if (   isNan (target)  // prediction must be large ??
@@ -5158,7 +5158,7 @@ void DistTree::qc () const
       {
         dissim. qc ();
         QC_IMPLY (! subDepth, dissim. target >= 0.0);
-        QC_IMPLY (! DistTree_sp::variance_min && ! subDepth && ! dissim. target, dissim. indiscernible ());
+        QC_IMPLY (/*! DistTree_sp::variance_min &&*/ ! subDepth && ! dissim. target, dissim. indiscernible ());
         leafSet. addUnique (Pair<const Leaf*> (dissim. leaf1, dissim. leaf2));
         if (subDepth)
           { QC_ASSERT (dissim. mult < INF); }
@@ -5220,7 +5220,7 @@ void DistTree::qc () const
     {
       const VectorPtr<Leaf>& leaves = it. second;
       QC_ASSERT (leaves. size () >= 2);
-      QC_ASSERT (! DistTree_sp::variance_min);
+    //QC_ASSERT (! DistTree_sp::variance_min);
       for (const Leaf* leaf : leaves)
       {
         QC_ASSERT (leaf->graph == this);
@@ -5560,7 +5560,7 @@ void DistTree::setDissimMult (bool usePrediction)
 
 
   // To keep absCriterion < INF
-  if (! multFixed && ! variance_min && usePrediction)
+  if (! multFixed /*&& ! variance_min*/ && usePrediction)
   {  
     unordered_map<const Leaf*,Real/*dissim.target*/> leaf2target_min;  
     leaf2target_min. rehash (name2leaf. size () / 100 + 1);  // PAR

phylogeny/distTree.hpp

@@ -652,7 +652,6 @@ struct Leaf : DTNode
   friend DissimLine;
   void collapse (Leaf* other);
     // Output: discernible = false
-    // Requires: !DistTree_sp::variance_min
     // Invokes: setParent()
     // To be followed by: DistTree::cleanTopology()
 public:	

phylogeny/distTree_new.cpp

@@ -63,7 +63,7 @@ struct ThisApplication : Application
   	  addKey ("variance", "Dissimilarity variance function: " + varianceTypeNames. toString (" | "), varianceTypeNames [varianceType]);
   	  addKey ("variance_power", "Power for -variance pow; >= 0", "NaN");
   	  addFlag ("variance_dissim", "Variance is computed off dissimilarities");
-  	  addKey ("variance_min", "Min. dissimilarity variance; to be added to the computed variance. If > 0 then all objects are discernible", "0");
+  	  addKey ("variance_min", "Min. dissimilarity variance; to be added to the computed variance", "0");
 
 		  addKey ("name", "Name of the object");
 		  addKey ("dissim", "File of the format: <obj1> <obj2> <dissimilarity>");

phylogeny/makeDistTree.cpp

@@ -76,7 +76,7 @@ struct ThisApplication : Application
 	  addKey ("variance_power", "Power for -variance pow; >= 0", "NaN");
 	  addFlag ("variance_dissim", "Variance is computed off dissimilarities");
 	  addFlag ("reinsert_variance_dist", "Variance is computed off tree distances for the reinsert optimization");
-	  addKey ("variance_min", "Min. dissimilarity variance. If > 0 then all objects are discernible", "0");  // ; to be added to the computed variance
+	  addKey ("variance_min", "Min. dissimilarity variance", "0");  // ; to be added to the computed variance
 
 	  // Processing
 	  addKey ("delete", "Delete leaves whose names are in the indicated file");

version.inc

@@ -1 +1 @@
-#define VERSION "1.1.7"
+#define VERSION "1.1.8"