/**
* Subspace relevance test.
*
* @param subspace Subspace to test
* @param neigh Neighbor list
* @param kernel Kernel density estimator
* @return relevance test result
*/
protected boolean relevantSubspace(
long[] subspace, DoubleDBIDList neigh, KernelDensityEstimator kernel) {
Relation<V> relation = kernel.relation;
final double crit = K_S_CRITICAL001 / Math.sqrt(neigh.size());
for (int dim = BitsUtil.nextSetBit(subspace, 0);
dim > 0;
dim = BitsUtil.nextSetBit(subspace, dim + 1)) {
// TODO: can we save this copy somehow?
double[] data = new double[neigh.size()];
{
int count = 0;
for (DBIDIter neighbor = neigh.iter(); neighbor.valid(); neighbor.advance()) {
V vector = relation.get(neighbor);
data[count] = vector.doubleValue(dim);
count++;
}
assert (count == neigh.size());
}
Arrays.sort(data);
final double norm = data[data.length - 1] - data[0];
final double min = data[0];
// Kolmogorow-Smirnow-Test against uniform distribution:
for (int j = 1; j < data.length - 2; j++) {
double delta = (j / (data.length - 1.)) - ((data[j] - min) / norm);
if (Math.abs(delta) > crit) {
return false;
}
}
}
return true;
}
/** Resets the values for the next cluster search. */
protected void reset() {
rows = BitsUtil.ones(rowM.length);
rowcard = rowM.length;
cols = BitsUtil.ones(colM.length);
colcard = colM.length;
BitsUtil.zeroI(irow);
}
/**
* Main loop for OUTRES
*
* @param relation Relation to process
* @return Outlier detection result
*/
public OutlierResult run(Relation<V> relation) {
WritableDoubleDataStore ranks =
DataStoreUtil.makeDoubleStorage(relation.getDBIDs(), DataStoreFactory.HINT_STATIC);
DoubleMinMax minmax = new DoubleMinMax();
KernelDensityEstimator kernel = new KernelDensityEstimator(relation);
long[] subspace = BitsUtil.zero(kernel.dim);
FiniteProgress progress =
LOG.isVerbose() ? new FiniteProgress("OUTRES scores", relation.size(), LOG) : null;
for (DBIDIter iditer = relation.iterDBIDs(); iditer.valid(); iditer.advance()) {
BitsUtil.zeroI(subspace);
double score = outresScore(0, subspace, iditer, kernel);
ranks.putDouble(iditer, score);
minmax.put(score);
LOG.incrementProcessed(progress);
}
LOG.ensureCompleted(progress);
OutlierScoreMeta meta =
new InvertedOutlierScoreMeta(minmax.getMin(), minmax.getMax(), 0., 1., 1.);
OutlierResult outresResult =
new OutlierResult(
meta,
new MaterializedDoubleRelation("OUTRES", "outres-score", ranks, relation.getDBIDs()));
return outresResult;
}
@Override
protected double minDistObject(SpatialComparable mbr, NumberVector v) {
if (mbr.getDimensionality() != v.getDimensionality()) {
throw new IllegalArgumentException(
"Different dimensionality of objects\n "
+ "first argument: "
+ mbr.toString()
+ "\n "
+ "second argument: "
+ v.toString());
}
double agg = 0.;
for (int d = BitsUtil.nextSetBit(dimensions, 0);
d >= 0;
d = BitsUtil.nextSetBit(dimensions, d + 1)) {
final double value = v.doubleValue(d);
final double omin = mbr.getMin(d);
final double diff1 = omin - value;
if (diff1 > 0.) {
if (diff1 > agg) {
agg = diff1;
}
} else {
final double omax = mbr.getMax(d);
final double diff2 = value - omax;
if (diff2 > agg) {
agg = diff2;
}
}
}
return agg;
}
public List<Polygon> compute() {
// Compute delaunay triangulation:
delaunay = (new SweepHullDelaunay2D(points)).getDelaunay();
List<Polygon> polys = new ArrayList<>();
// Working data
long[] used = BitsUtil.zero(delaunay.size());
List<double[]> cur = new ArrayList<>();
for (int i = 0 /* = used.nextClearBit(0) */;
i < delaunay.size() && i >= 0;
i = BitsUtil.nextClearBit(used, i + 1)) {
if (!BitsUtil.get(used, i)) {
BitsUtil.setI(used, i);
SweepHullDelaunay2D.Triangle tri = delaunay.get(i);
if (tri.r2 <= alpha2) {
// Check neighbors
processNeighbor(cur, used, i, tri.ab, tri.b);
processNeighbor(cur, used, i, tri.bc, tri.c);
processNeighbor(cur, used, i, tri.ca, tri.a);
}
if (cur.size() > 0) {
polys.add(new Polygon(cur));
cur = new ArrayList<>();
}
}
}
return polys;
}
@Override
public double minDist(SpatialComparable mbr1, SpatialComparable mbr2) {
if (mbr1.getDimensionality() != mbr2.getDimensionality()) {
throw new IllegalArgumentException(
"Different dimensionality of objects\n "
+ "first argument: "
+ mbr1.toString()
+ "\n "
+ "second argument: "
+ mbr2.toString());
}
double agg = 0.;
for (int d = BitsUtil.nextSetBit(dimensions, 0);
d >= 0;
d = BitsUtil.nextSetBit(dimensions, d + 1)) {
final double max1 = mbr1.getMax(d);
final double min2 = mbr2.getMin(d);
if (max1 < min2) {
double v = min2 - max1;
if (v > agg) {
agg = v;
}
} else {
final double min1 = mbr1.getMin(d);
final double max2 = mbr2.getMax(d);
double v = min1 - max2;
if (v > agg) {
agg = v;
}
}
}
return agg;
}
/**
* Select or deselect a column.
*
* @param cnum Column to select
* @param set Value to set
*/
protected void selectColumn(int cnum, boolean set) {
if (set) {
BitsUtil.setI(cols, cnum);
colcard++;
} else {
BitsUtil.clearI(cols, cnum);
colcard--;
}
}
/**
* Select or deselect a row.
*
* @param rnum Row to select
* @param set Value to set
*/
protected void selectRow(int rnum, boolean set) {
if (set) {
BitsUtil.setI(rows, rnum);
rowcard++;
} else {
BitsUtil.clearI(rows, rnum);
rowcard--;
}
}
/**
* Constructor.
*
* @param rows Row dimensionality.
* @param cols Column dimensionality.
*/
protected BiclusterCandidate(int rows, int cols) {
super();
this.rows = BitsUtil.ones(rows);
this.irow = BitsUtil.zero(rows);
this.rowcard = rows;
this.rowM = new double[rows];
this.cols = BitsUtil.ones(cols);
this.colcard = cols;
this.colM = new double[cols];
}
@Override
public double norm(NumberVector obj) {
double agg = 0.;
for (int d = BitsUtil.nextSetBit(dimensions, 0);
d >= 0;
d = BitsUtil.nextSetBit(dimensions, d + 1)) {
double v = Math.abs(obj.doubleValue(d));
if (v > agg) {
agg = v;
}
}
return agg;
}
/**
* Visit a column of the matrix.
*
* @param mat Data matrix
* @param col Column to visit
* @param mode Operation mode
* @param visitor Visitor function
*/
protected void visitColumn(double[][] mat, int col, int mode, CellVisitor visitor) {
boolean cselected = BitsUtil.get(cols, col);
// For efficiency, we manually iterate over the rows and column bitmasks.
// This saves repeated shifting needed by the manual bit access.
for (int rpos = 0, rlpos = 0; rlpos < rows.length; ++rlpos) {
long rlong = rows[rlpos];
// Fast skip blocks of 64 masked values.
if (mode == CellVisitor.SELECTED && rlong == 0L) {
rpos += Long.SIZE;
continue;
}
if (mode == CellVisitor.NOT_SELECTED && rlong == -1L) {
rpos += Long.SIZE;
continue;
}
for (int i = 0; i < Long.SIZE && rpos < rowM.length; ++i, ++rpos, rlong >>>= 1) {
boolean rselected = ((rlong & 1L) == 1L);
if (mode == CellVisitor.SELECTED && !rselected) {
continue;
}
if (mode == CellVisitor.NOT_SELECTED && rselected) {
continue;
}
boolean stop = visitor.visit(mat[rpos][col], rpos, col, rselected, cselected);
if (stop) {
return;
}
}
}
}
/**
* Visit a row of the data matrix.
*
* @param mat Data matrix
* @param row Row to visit
* @param visitor Visitor function
*/
protected void visitRow(double[][] mat, int row, int mode, CellVisitor visitor) {
boolean rselected = BitsUtil.get(rows, row);
final double[] rowdata = mat[row];
for (int cpos = 0, clpos = 0; clpos < cols.length; ++clpos) {
long clong = cols[clpos];
// Fast skip blocks of 64 masked values.
if (mode == CellVisitor.SELECTED && clong == 0L) {
cpos += Long.SIZE;
continue;
}
if (mode == CellVisitor.NOT_SELECTED && clong == -1L) {
cpos += Long.SIZE;
continue;
}
for (int j = 0; j < Long.SIZE && cpos < colM.length; ++j, ++cpos, clong >>>= 1) {
boolean cselected = ((clong & 1L) == 1L);
if (mode == CellVisitor.SELECTED && !cselected) {
continue;
}
if (mode == CellVisitor.NOT_SELECTED && cselected) {
continue;
}
boolean stop = visitor.visit(rowdata[cpos], row, cpos, rselected, cselected);
if (stop) {
return;
}
}
}
}
@Override
public long[] getVisibleDimensions2D() {
final int dim = proj.getDimensionality();
long[] actDim = BitsUtil.zero(dim);
double[] vScale = new double[dim];
for (int d = 0; d < dim; d++) {
Arrays.fill(vScale, 0);
vScale[d] = 1;
double[] vRender = fastProjectScaledToRenderSpace(vScale);
// TODO: Can't we do this by inspecting the projection matrix directly?
if (vRender[0] > 0.0 || vRender[0] < 0.0 || vRender[1] != 0) {
BitsUtil.setI(actDim, d);
}
}
return actDim;
}
/**
* Main loop of OUTRES. Run for each object
*
* @param s start dimension
* @param subspace Current subspace
* @param id Current object ID
* @param kernel Kernel
* @return Score
*/
public double outresScore(
final int s, long[] subspace, DBIDRef id, KernelDensityEstimator kernel) {
double score = 1.0; // Initial score is 1.0
final SubspaceEuclideanDistanceFunction df = new SubspaceEuclideanDistanceFunction(subspace);
MeanVariance meanv = new MeanVariance();
for (int i = s; i < kernel.dim; i++) {
if (BitsUtil.get(subspace, i)) { // TODO: needed? Or should we always start
// with i=0?
continue;
}
BitsUtil.setI(subspace, i);
df.setSelectedDimensions(subspace);
final double adjustedEps = kernel.adjustedEps(kernel.dim);
// Query with a larger window, to also get neighbors of neighbors
// Subspace euclidean is metric!
final double range = adjustedEps * 2.;
RangeQuery<V> rq = QueryUtil.getRangeQuery(kernel.relation, df, range);
DoubleDBIDList neighc = rq.getRangeForDBID(id, range);
DoubleDBIDList neigh = refineRange(neighc, adjustedEps);
if (neigh.size() > 2) {
// Relevance test
if (relevantSubspace(subspace, neigh, kernel)) {
final double density = kernel.subspaceDensity(subspace, neigh);
// Compute mean and standard deviation for densities of neighbors.
meanv.reset();
for (DoubleDBIDListIter neighbor = neigh.iter(); neighbor.valid(); neighbor.advance()) {
DoubleDBIDList n2 = subsetNeighborhoodQuery(neighc, neighbor, df, adjustedEps, kernel);
meanv.put(kernel.subspaceDensity(subspace, n2));
}
final double deviation = (meanv.getMean() - density) / (2. * meanv.getSampleStddev());
// High deviation:
if (deviation >= 1) {
score *= (density / deviation);
}
// Recursion
score *= outresScore(i + 1, subspace, id, kernel);
}
}
BitsUtil.clearI(subspace, i);
}
return score;
}
@Override
public Clustering<BiclusterWithInversionsModel> biclustering() {
double[][] mat = RelationUtil.relationAsMatrix(relation, rowIDs);
BiclusterCandidate cand = new BiclusterCandidate(getRowDim(), getColDim());
Clustering<BiclusterWithInversionsModel> result =
new Clustering<>("Cheng-and-Church", "Cheng and Church Biclustering");
ModifiableDBIDs noise = DBIDUtil.newHashSet(relation.getDBIDs());
FiniteProgress prog = LOG.isVerbose() ? new FiniteProgress("Extracting Cluster", n, LOG) : null;
for (int i = 0; i < n; i++) {
cand.reset();
multipleNodeDeletion(mat, cand);
if (LOG.isVeryVerbose()) {
LOG.veryverbose(
"Residue after Alg 2: " + cand.residue + " " + cand.rowcard + "x" + cand.colcard);
}
singleNodeDeletion(mat, cand);
if (LOG.isVeryVerbose()) {
LOG.veryverbose(
"Residue after Alg 1: " + cand.residue + " " + cand.rowcard + "x" + cand.colcard);
}
nodeAddition(mat, cand);
if (LOG.isVeryVerbose()) {
LOG.veryverbose(
"Residue after Alg 3: " + cand.residue + " " + cand.rowcard + "x" + cand.colcard);
}
cand.maskMatrix(mat, dist);
BiclusterWithInversionsModel model =
new BiclusterWithInversionsModel(colsBitsetToIDs(cand.cols), rowsBitsetToIDs(cand.irow));
final ArrayDBIDs cids = rowsBitsetToIDs(cand.rows);
noise.removeDBIDs(cids);
result.addToplevelCluster(new Cluster<>(cids, model));
if (LOG.isVerbose()) {
LOG.verbose("Score of bicluster " + (i + 1) + ": " + cand.residue + "\n");
LOG.verbose("Number of rows: " + cand.rowcard + "\n");
LOG.verbose("Number of columns: " + cand.colcard + "\n");
// LOG.verbose("Total number of masked values: " + maskedVals.size() +
// "\n");
}
LOG.incrementProcessed(prog);
}
// Add a noise cluster, full-dimensional.
if (!noise.isEmpty()) {
long[] allcols = BitsUtil.ones(getColDim());
BiclusterWithInversionsModel model =
new BiclusterWithInversionsModel(colsBitsetToIDs(allcols), DBIDUtil.EMPTYDBIDS);
result.addToplevelCluster(new Cluster<>(noise, true, model));
}
LOG.ensureCompleted(prog);
return result;
}
/**
* Compute density in the given subspace.
*
* @param subspace Subspace
* @param neighbors Neighbor distance list
* @return Density
*/
protected double subspaceDensity(long[] subspace, DoubleDBIDList neighbors) {
final double bandwidth = optimalBandwidth(BitsUtil.cardinality(subspace));
double density = 0;
for (DoubleDBIDListIter neighbor = neighbors.iter(); neighbor.valid(); neighbor.advance()) {
double v = neighbor.doubleValue() / bandwidth;
if (v < 1) {
density += 1 - (v * v);
}
}
return density / relation.size();
}
@Override
public double distance(NumberVector v1, NumberVector v2) {
if (v1.getDimensionality() != v2.getDimensionality()) {
throw new IllegalArgumentException(
"Different dimensionality of FeatureVectors\n "
+ "first argument: "
+ v1
+ "\n "
+ "second argument: "
+ v2);
}
double agg = 0.;
for (int d = BitsUtil.nextSetBit(dimensions, 0);
d >= 0;
d = BitsUtil.nextSetBit(dimensions, d + 1)) {
double v = Math.abs(v1.doubleValue(d) - v2.doubleValue(d));
if (v > agg) {
agg = v;
}
}
return agg;
}
private void processNeighbor(List<double[]> cur, long[] used, int i, int ab, int b) {
if (ab >= 0) {
if (BitsUtil.get(used, ab)) {
return;
}
BitsUtil.setI(used, ab);
final SweepHullDelaunay2D.Triangle next = delaunay.get(ab);
if (next.r2 < alpha2) {
// Continue where we left off...
if (next.ab == i) {
processNeighbor(cur, used, ab, next.bc, next.c);
processNeighbor(cur, used, ab, next.ca, next.a);
} else if (next.bc == i) {
processNeighbor(cur, used, ab, next.ca, next.a);
processNeighbor(cur, used, ab, next.ab, next.b);
} else if (next.ca == i) {
processNeighbor(cur, used, ab, next.ab, next.b);
processNeighbor(cur, used, ab, next.bc, next.c);
}
return;
}
}
cur.add(points.get(b));
}
protected void invertRow(int rnum, boolean b) {
BitsUtil.setI(irow, rnum);
}
/**
* Performs the DOC or FastDOC (as configured) algorithm on the given Database.
*
* <p>This will run exhaustively, i.e. run DOC until no clusters are found anymore / the database
* size has shrunk below the threshold for minimum cluster size.
*
* @param database Database
* @param relation Data relation
*/
public Clustering<SubspaceModel> run(Database database, Relation<V> relation) {
// Dimensionality of our set.
final int d = RelationUtil.dimensionality(relation);
// Get available DBIDs as a set we can remove items from.
ArrayModifiableDBIDs S = DBIDUtil.newArray(relation.getDBIDs());
// Precompute values as described in Figure 2.
double r = Math.abs(Math.log(d + d) / Math.log(beta * .5));
// Outer loop count.
int n = (int) (2. / alpha);
// Inner loop count.
int m = (int) (Math.pow(2. / alpha, r) * Math.log(4));
if (heuristics) {
m = Math.min(m, Math.min(1000000, d * d));
}
// Minimum size for a cluster for it to be accepted.
int minClusterSize = (int) (alpha * S.size());
// List of all clusters we found.
Clustering<SubspaceModel> result = new Clustering<>("DOC Clusters", "DOC");
// Inform the user about the number of actual clusters found so far.
IndefiniteProgress cprogress =
LOG.isVerbose() ? new IndefiniteProgress("Number of clusters", LOG) : null;
// To not only find a single cluster, we continue running until our set
// of points is empty.
while (S.size() > minClusterSize) {
Cluster<SubspaceModel> C;
if (heuristics) {
C = runFastDOC(database, relation, S, d, n, m, (int) r);
} else {
C = runDOC(database, relation, S, d, n, m, (int) r, minClusterSize);
}
if (C == null) {
// Stop trying if we couldn't find a cluster.
break;
}
// Found a cluster, remember it, remove its points from the set.
result.addToplevelCluster(C);
// Remove all points of the cluster from the set and continue.
S.removeDBIDs(C.getIDs());
if (cprogress != null) {
cprogress.setProcessed(result.getAllClusters().size(), LOG);
}
}
// Add the remainder as noise.
if (S.size() > 0) {
long[] alldims = BitsUtil.ones(d);
result.addToplevelCluster(
new Cluster<>(
S,
true,
new SubspaceModel(new Subspace(alldims), Centroid.make(relation, S).getArrayRef())));
}
LOG.setCompleted(cprogress);
return result;
}
/**
* Performs a single run of DOC, finding a single cluster.
*
* @param database Database context
* @param relation used to get actual values for DBIDs.
* @param S The set of points we're working on.
* @param d Dimensionality of the data set we're currently working on.
* @param r Size of random samples.
* @param m Number of inner iterations (per seed point).
* @param n Number of outer iterations (seed points).
* @param minClusterSize Minimum size a cluster must have to be accepted.
* @return a cluster, if one is found, else <code>null</code>.
*/
private Cluster<SubspaceModel> runDOC(
Database database,
Relation<V> relation,
ArrayModifiableDBIDs S,
final int d,
int n,
int m,
int r,
int minClusterSize) {
// Best cluster for the current run.
DBIDs C = null;
// Relevant attributes for the best cluster.
long[] D = null;
// Quality of the best cluster.
double quality = Double.NEGATIVE_INFINITY;
// Bounds for our cluster.
// ModifiableHyperBoundingBox bounds = new ModifiableHyperBoundingBox(new
// double[d], new double[d]);
// Weights for distance (= rectangle query)
SubspaceMaximumDistanceFunction df = new SubspaceMaximumDistanceFunction(BitsUtil.zero(d));
DistanceQuery<V> dq = database.getDistanceQuery(relation, df);
RangeQuery<V> rq = database.getRangeQuery(dq);
// Inform the user about the progress in the current iteration.
FiniteProgress iprogress =
LOG.isVerbose()
? new FiniteProgress("Iteration progress for current cluster", m * n, LOG)
: null;
Random random = rnd.getSingleThreadedRandom();
DBIDArrayIter iter = S.iter();
for (int i = 0; i < n; ++i) {
// Pick a random seed point.
iter.seek(random.nextInt(S.size()));
for (int j = 0; j < m; ++j) {
// Choose a set of random points.
DBIDs randomSet = DBIDUtil.randomSample(S, r, random);
// Initialize cluster info.
long[] nD = BitsUtil.zero(d);
// Test each dimension and build bounding box.
for (int k = 0; k < d; ++k) {
if (dimensionIsRelevant(k, relation, randomSet)) {
BitsUtil.setI(nD, k);
}
}
if (BitsUtil.cardinality(nD) > 0) {
// Get all points in the box.
df.setSelectedDimensions(nD);
// TODO: add filtering capabilities into query API!
DBIDs nC = DBIDUtil.intersection(S, rq.getRangeForDBID(iter, w));
if (LOG.isDebuggingFiner()) {
LOG.finer(
"Testing a cluster candidate, |C| = "
+ nC.size()
+ ", |D| = "
+ BitsUtil.cardinality(nD));
}
// Is the cluster large enough?
if (nC.size() < minClusterSize) {
// Too small.
if (LOG.isDebuggingFiner()) {
LOG.finer("... but it's too small.");
}
} else {
// Better cluster than before?
double nQuality = computeClusterQuality(nC.size(), BitsUtil.cardinality(nD));
if (nQuality > quality) {
if (LOG.isDebuggingFiner()) {
LOG.finer("... and it's the best so far: " + nQuality + " vs. " + quality);
}
C = nC;
D = nD;
quality = nQuality;
} else {
if (LOG.isDebuggingFiner()) {
LOG.finer("... but we already have a better one.");
}
}
}
}
LOG.incrementProcessed(iprogress);
}
}
LOG.ensureCompleted(iprogress);
return (C != null) ? makeCluster(relation, C, D) : null;
}
/**
* Performs a single run of FastDOC, finding a single cluster.
*
* @param database Database context
* @param relation used to get actual values for DBIDs.
* @param S The set of points we're working on.
* @param d Dimensionality of the data set we're currently working on.
* @param r Size of random samples.
* @param m Number of inner iterations (per seed point).
* @param n Number of outer iterations (seed points).
* @return a cluster, if one is found, else <code>null</code>.
*/
private Cluster<SubspaceModel> runFastDOC(
Database database, Relation<V> relation, ArrayModifiableDBIDs S, int d, int n, int m, int r) {
// Relevant attributes of highest cardinality.
long[] D = null;
// The seed point for the best dimensions.
DBIDVar dV = DBIDUtil.newVar();
// Inform the user about the progress in the current iteration.
FiniteProgress iprogress =
LOG.isVerbose()
? new FiniteProgress("Iteration progress for current cluster", m * n, LOG)
: null;
Random random = rnd.getSingleThreadedRandom();
DBIDArrayIter iter = S.iter();
outer:
for (int i = 0; i < n; ++i) {
// Pick a random seed point.
iter.seek(random.nextInt(S.size()));
for (int j = 0; j < m; ++j) {
// Choose a set of random points.
DBIDs randomSet = DBIDUtil.randomSample(S, r, random);
// Initialize cluster info.
long[] nD = BitsUtil.zero(d);
// Test each dimension.
for (int k = 0; k < d; ++k) {
if (dimensionIsRelevant(k, relation, randomSet)) {
BitsUtil.setI(nD, k);
}
}
if (D == null || BitsUtil.cardinality(nD) > BitsUtil.cardinality(D)) {
D = nD;
dV.set(iter);
if (BitsUtil.cardinality(D) >= d_zero) {
if (iprogress != null) {
iprogress.setProcessed(iprogress.getTotal(), LOG);
}
break outer;
}
}
LOG.incrementProcessed(iprogress);
}
}
LOG.ensureCompleted(iprogress);
// If no relevant dimensions were found, skip it.
if (D == null || BitsUtil.cardinality(D) == 0) {
return null;
}
// Get all points in the box.
SubspaceMaximumDistanceFunction df = new SubspaceMaximumDistanceFunction(D);
DistanceQuery<V> dq = database.getDistanceQuery(relation, df);
RangeQuery<V> rq = database.getRangeQuery(dq, DatabaseQuery.HINT_SINGLE);
// TODO: add filtering capabilities into query API!
DBIDs C = DBIDUtil.intersection(S, rq.getRangeForDBID(dV, w));
// If we have a non-empty cluster, return it.
return (C.size() > 0) ? makeCluster(relation, C, D) : null;
}
