import java.util.ArrayList;
import java.util.List;
import java.util.Map.Entry;
-import java.util.SortedMap;
import java.util.SortedSet;
-import java.util.TreeMap;
-import java.util.TreeSet;
import org.forester.evoinference.matrix.distance.BasicSymmetricalDistanceMatrix;
import org.forester.phylogeny.Phylogeny;
public final class NeighborJoiningR {
- private BasicSymmetricalDistanceMatrix _d;
- private double[][] _d_values;
- private final DecimalFormat _df;
- private PhylogenyNode[] _external_nodes;
- private int[] _mappings;
- private int _n;
- private double[] _r;
- private final boolean _verbose;
- private int _min_i;
- private int _min_j;
- private List<SortedMap<Double, SortedSet<Integer>>> _s;
+ private final static DecimalFormat DF = new DecimalFormat( "0.00000" );
+ private BasicSymmetricalDistanceMatrix _d;
+ private double[][] _d_values;
+ private final DecimalFormat _df;
+ private PhylogenyNode[] _external_nodes;
+ private int[] _mappings;
+ private int _n;
+ private double[] _r;
+ private final boolean _verbose;
+ private int _min_i;
+ private int _min_j;
+ private S _s;
+ private double _d_min; //TODO remove me
+ private int[] _rev_mappings;
private NeighborJoiningR() {
_verbose = false;
reset( distance );
final Phylogeny phylogeny = new Phylogeny();
while ( _n > 2 ) {
+ System.out.println( "N=" + _n );
+ System.out.println();
// Calculates the minimal distance.
// If more than one minimal distances, always the first found is used
- updateM();
+ final double m = updateM();
final int otu1 = _min_i;
final int otu2 = _min_j;
+ System.out.println( _min_i + " " + _min_j + " => " + DF.format( m ) + " (" + DF.format( _d_min ) + ")" );
// It is a condition that otu1 < otu2.
+ //System.out.println( "mapped 1 " + _mappings[ otu1 ] );
+ System.out.println( "mapped otu 2 " + _mappings[ otu2 ] );
final PhylogenyNode node = new PhylogenyNode();
- final double d = getDvalue( otu1, otu2 );
- final double d1 = ( d / 2 ) + ( ( _r[ otu1 ] - _r[ otu2 ] ) / ( 2 * ( _n - 2 ) ) );
+ //final double d = getDvalueUnmapped( otu1, _mappings[ otu2 ] );
+ final double d = _d_values[ otu1 ][ _mappings[ otu2 ] ];
+ final double d1 = ( d / 2 ) + ( ( _r[ _rev_mappings[ otu1 ] ] - _r[ otu2 ] ) / ( 2 * ( _n - 2 ) ) );
final double d2 = d - d1;
if ( _df == null ) {
- getExternalPhylogenyNode( otu1 ).setDistanceToParent( d1 );
+ _external_nodes[ otu1 ].setDistanceToParent( d1 );
getExternalPhylogenyNode( otu2 ).setDistanceToParent( d2 );
}
else {
// yes, yes, slow but only grows with n (and not n^2 or worse)...
- getExternalPhylogenyNode( otu1 ).setDistanceToParent( Double.parseDouble( _df.format( d1 ) ) );
+ _external_nodes[ otu1 ].setDistanceToParent( Double.parseDouble( _df.format( d1 ) ) );
getExternalPhylogenyNode( otu2 ).setDistanceToParent( Double.parseDouble( _df.format( d2 ) ) );
}
- node.addAsChild( getExternalPhylogenyNode( otu1 ) );
+ node.addAsChild( _external_nodes[ otu1 ] );
node.addAsChild( getExternalPhylogenyNode( otu2 ) );
if ( _verbose ) {
- printProgress( otu1, otu2 );
+ printProgress( otu1, otu2, node );
}
+ System.out.println( "otu1=" + otu1 );
+ System.out.println( "otu2=" + otu2 );
calculateDistancesFromNewNode( otu1, otu2, d );
- _external_nodes[ _mappings[ otu1 ] ] = node;
+ // _external_nodes[ _mappings[ otu1 ] ] = node;
+ _external_nodes[ otu1 ] = node;
updateMappings( otu2 );
--_n;
+ System.out.println( "" );
+ System.out.println( "----------------------------------------------------------------------------------" );
+ System.out.println( "" );
}
final double d = getDvalue( 0, 1 ) / 2;
if ( _df == null ) {
root.addAsChild( getExternalPhylogenyNode( 0 ) );
root.addAsChild( getExternalPhylogenyNode( 1 ) );
if ( _verbose ) {
- printProgress( 0, 1 );
+ printProgress( 0, 1, root );
}
phylogeny.setRoot( root );
phylogeny.setRooted( false );
}
private final void calculateDistancesFromNewNode( final int otu1, final int otu2, final double d ) {
- for( int i = 0; i < _n; ++i ) {
- if ( ( i == otu1 ) || ( i == otu2 ) ) {
+ System.out.print( "new D values: " );
+ for( int j = 0; j < _n; ++j ) {
+ if ( j == otu2 ) {
continue;
}
- updateDvalue( otu1, otu2, i, d );
+ if ( otu1 < _mappings[ j ] ) {
+ updateDvalue( otu1, otu2, j, d );
+ }
}
+ System.out.println();
}
- private final void updateDvalue( final int otu1, final int otu2, final int i, final double d ) {
- setDvalue( otu1, i, ( getDvalue( otu1, i ) + getDvalue( i, otu2 ) - d ) / 2 );
+ private final void updateDvalue( final int otu1, final int otu2, final int j, final double d ) {
+ final double new_d = ( getDvalueUnmapped( otu1, _mappings[ j ] ) + getDvalue( j, otu2 ) - d ) / 2;
+ System.out.print( DF.format( new_d ) + " " );
+ // System.out.println( "going to remove: " + getDvalueUnmapped( otu1, _mappings[ j ] ) + ", " + otu1 + ", "
+ // + _mappings[ j ] );
+
+ if ( otu1< _mappings[ j ] ) {
+ _s.removePairing( getDvalueUnmapped( otu1, _mappings[ j ] ), otu1, _mappings[ j ] );
+ }
+ else {
+ _s.removePairing( getDvalueUnmapped( otu1, _mappings[ j ] ), _mappings[ j ] , otu1 );
+ }
+
+
+ // System.out.println( "going to remove: " + getDvalue( j, otu2 ) + ", " +_mappings[ otu2 ] + ", "
+ // + _mappings[ j ] );
+
+ if ( _mappings[ otu2 ] < _mappings[ j ] ) {
+ _s.removePairing( getDvalue( j, otu2 ), _mappings[ otu2 ] , _mappings[ j ] );
+ }
+ else {
+ _s.removePairing( getDvalue( j, otu2 ) , _mappings[ j ], _mappings[ otu2 ] );
+
+ }
+
+ _s.addPairing( new_d, otu1, _mappings[ j ] );
+ setDvalueU( otu1, j, new_d );
}
- private void setDvalue( final int i, final int j, final double d ) {
- if ( i < j ) {
- _d_values[ _mappings[ i ] ][ _mappings[ j ] ] = d;
+ private void setDvalueU( final int i, final int j, final double d ) {
+ if ( i < _mappings[ j ] ) {
+ _d_values[ i ][ _mappings[ j ] ] = d;
}
- _d_values[ _mappings[ j ] ][ _mappings[ i ] ] = d;
+ _d_values[_mappings[ j] ][ i ] = d;
}
private double getDvalue( final int i, final int j ) {
return _d_values[ _mappings[ j ] ][ _mappings[ i ] ];
}
+ private double getDvalueUnmapped( final int i, final int j ) {
+ if ( i < j ) {
+ return _d_values[ i ][ j ];
+ }
+ return _d_values[ j ][ i ];
+ }
+
private final void calculateNetDivergences() {
for( int i = 0; i < _n; ++i ) {
_r[ i ] = calculateNetDivergence( i );
_external_nodes[ i ].setName( Integer.toString( i ) );
}
_mappings[ i ] = i;
+ _rev_mappings[ i ] = i;
}
}
- private final void printProgress( final int otu1, final int otu2 ) {
- final PhylogenyNode n1 = getExternalPhylogenyNode( otu1 );
- final PhylogenyNode n2 = getExternalPhylogenyNode( otu2 );
- System.out.println( "Node " + ( ForesterUtil.isEmpty( n1.getName() ) ? n1.getId() : n1.getName() ) + " joins "
- + ( ForesterUtil.isEmpty( n2.getName() ) ? n2.getId() : n2.getName() ) );
+ private final void printProgress( final int otu1, final int otu2, final PhylogenyNode node ) {
+ System.out.println( "Node " + printProgressNodeToString( _external_nodes[ otu1 ] ) + " joins "
+ + ( printProgressNodeToString( getExternalPhylogenyNode( otu2 ) ) ) + " [resulting in node "
+ + ( printProgressNodeToString( node ) ) + "]" );
+ }
+
+ private final String printProgressNodeToString( final PhylogenyNode n ) {
+ if ( n.isExternal() ) {
+ if ( ForesterUtil.isEmpty( n.getName() ) ) {
+ return Long.toString( n.getId() );
+ }
+ return n.getName();
+ }
+ return n.getId()
+ + " ("
+ + ( ForesterUtil.isEmpty( n.getChildNode1().getName() ) ? n.getChildNode1().getId() : n.getChildNode1()
+ .getName() )
+ + "+"
+ + ( ForesterUtil.isEmpty( n.getChildNode2().getName() ) ? n.getChildNode2().getId() : n.getChildNode2()
+ .getName() ) + ")";
}
// only the values in the lower triangle are used.
_d = distances;
_r = new double[ _n ];
_mappings = new int[ _n ];
+ _rev_mappings = new int[ _n ];
_d_values = _d.getValues();
- _s = new ArrayList<SortedMap<Double, SortedSet<Integer>>>();
- for( int j = 0; j < _n; ++j ) {
- final TreeMap<Double, SortedSet<Integer>> map = new TreeMap<Double, SortedSet<Integer>>();
- for( int i = 0; i < j; ++i ) {
- if ( !map.containsKey( _d_values[ i ][ j ] ) ) {
- map.put( _d_values[ i ][ j ], new TreeSet<Integer>() );
- }
- map.get( _d_values[ i ][ j ] ).add( i );
- }
- _s.add( map );
- }
+ _s = new S();
+ _s.initialize( distances );
initExternalNodes();
+ System.out.println();
printM();
+ System.out.println( "----------------------------------------------------------------------------------" );
+ System.out.println();
+ System.out.println();
}
final private void printM() {
- for( int j = 1; j < _n; ++j ) {
+ for( int j = 0; j < _d_values.length; ++j ) {
+ System.out.print( _external_nodes[ j ] );
+ System.out.print( "\t\t" );
+ for( int i = 0; i < _d_values[ j ].length; ++i ) {
+ System.out.print( DF.format( _d_values[ i ][ j ] ) );
+ System.out.print( " " );
+ }
+ System.out.println();
+ }
+ for( int j = 0; j < _n; ++j ) {
+ System.out.print( getExternalPhylogenyNode( j ) );
+ System.out.print( "\t\t" );
for( int i = 0; i < _n; ++i ) {
- System.out.print( _d_values[ i ][ j ] );
+ System.out.print( DF.format( _d_values[ _mappings[ i ] ][ _mappings[ j ] ] ) );
System.out.print( " " );
}
- System.out.print( " " );
- for( final Entry<Double, SortedSet<Integer>> entry : _s.get( j ).entrySet() ) {
- final double key = entry.getKey();
- final SortedSet<Integer> value = entry.getValue();
- System.out.print( key + "=" );
+ System.out.print( "\t\t" );
+ for( final Entry<Integer, SortedSet<Integer>> entry : _s.getSentrySet( _mappings[ j ] ) ) {
+ System.out.print( DF.format( ( double ) entry.getKey() / S.FACTOR ) + "=" );
boolean first = true;
- for( final Integer v : value ) {
+ for( final int v : entry.getValue() ) {
if ( !first ) {
System.out.print( "," );
}
}
}
- private final void updateM() {
+ private final double updateM() {
calculateNetDivergences();
- Double min = Double.MAX_VALUE;
+ Double min_m = Double.MAX_VALUE;
_min_i = -1;
_min_j = -1;
final int n_minus_2 = _n - 2;
+ printM();
for( int j = 1; j < _n; ++j ) {
final double r_j = _r[ j ];
final int m_j = _mappings[ j ];
- final SortedMap<Double, SortedSet<Integer>> s_j = _s.get( m_j );
- for( final Entry<Double, SortedSet<Integer>> entry : s_j.entrySet() ) {
- //Double key = entry.getKey();
- final SortedSet<Integer> value = entry.getValue();
- for( final Integer sorted_i : value ) {
+ System.out.print( "j=" + j + " mj=" + m_j + ": " );
+ for( final Entry<Integer, SortedSet<Integer>> entry : _s.getSentrySet( m_j ) ) {
+ for( final int sorted_i : entry.getValue() ) {
System.out.print( sorted_i + " " );
- // final double m = _d_values[ _mappings[ sorted_i ] ][ m_j ]
- // - ( ( _r[ sorted_i ] + r_j ) / n_minus_2 );
- // if ( m < min ) {
- // min = m;
- // _min_i = sorted_i;
- // _min_j = j;
- // }
+ System.out.print( "(" + DF.format( getDvalueUnmapped( sorted_i, m_j ) ) + ") " );
+ final double m = getDvalueUnmapped( sorted_i, m_j )
+ - ( ( _r[ _rev_mappings[ sorted_i ] ] + r_j ) / n_minus_2 );
+ if ( ( m < min_m ) ) {
+ _d_min = getDvalueUnmapped( sorted_i, m_j );
+ min_m = m;
+ _min_i = sorted_i;
+ _min_j = j;
+ }
+ }
+ }
+ System.out.println();
+ for( final Entry<Integer, SortedSet<Integer>> entry : _s.getSentrySet( m_j ) ) {
+ for( final int sorted_i : entry.getValue() ) {
+ System.out.print( sorted_i );
+ System.out.print( "->" );
+ System.out.print( DF.format( _r[ sorted_i ] ) );
+ System.out.print( " " );
}
}
System.out.println();
+ /*
for( int i = 0; i < j; ++i ) {
final double m = getDvalue( i, j ) - ( ( _r[ i ] + r_j ) / n_minus_2 );
if ( m < min ) {
min = m;
+ _d_min = getDvalue( i, j );
_min_i = i;
_min_j = j;
}
- }
+ }*/
}
+ System.out.println();
+ return min_m;
}
// otu2 will, in effect, be "deleted" from the matrix.
private final void updateMappings( final int otu2 ) {
for( int i = otu2; i < ( _mappings.length - 1 ); ++i ) {
+ System.out.print( _mappings[ i ] );
_mappings[ i ] = _mappings[ i + 1 ];
+ System.out.println( "----->" + _mappings[ i ] );
+ }
+ for( int i = 0; i < _mappings.length; ++i ) {
+ System.out.println( i + "-->" + _mappings[ i ] );
+ }
+ for( int i = 0; i < _n; ++i ) {
+ _rev_mappings[ _mappings[ i ] ] = i;
}
}
git://source.jalview.org / jalview.git / blobdiff