use of smile.clustering.linkage.UPGMCLinkage in project smile by haifengl.
the class HierarchicalClusteringTest method testUSPS.
/**
* Test of learn method, of class GMeans.
*/
@Test
public void testUSPS() {
System.out.println("USPS");
DelimitedTextParser parser = new DelimitedTextParser();
parser.setResponseIndex(new NominalAttribute("class"), 0);
try {
AttributeDataset train = parser.parse("USPS Train", smile.data.parser.IOUtils.getTestDataFile("usps/zip.train"));
double[][] x = train.toArray(new double[train.size()][]);
int[] y = train.toArray(new int[train.size()]);
int n = x.length;
double[][] proximity = new double[n][];
for (int i = 0; i < n; i++) {
proximity[i] = new double[i + 1];
for (int j = 0; j < i; j++) {
proximity[i][j] = Math.distance(x[i], x[j]);
}
}
AdjustedRandIndex ari = new AdjustedRandIndex();
RandIndex rand = new RandIndex();
HierarchicalClustering hc = new HierarchicalClustering(new SingleLinkage(proximity));
int[] label = hc.partition(10);
double r = rand.measure(y, label);
double r2 = ari.measure(y, label);
System.out.format("SingleLinkage rand index = %.2f%%\tadjusted rand index = %.2f%%%n", 100.0 * r, 100.0 * r2);
assertTrue(r > 0.1);
hc = new HierarchicalClustering(new CompleteLinkage(proximity));
label = hc.partition(10);
r = rand.measure(y, label);
r2 = ari.measure(y, label);
System.out.format("CompleteLinkage rand index = %.2f%%\tadjusted rand index = %.2f%%%n", 100.0 * r, 100.0 * r2);
assertTrue(r > 0.75);
hc = new HierarchicalClustering(new UPGMALinkage(proximity));
label = hc.partition(10);
r = rand.measure(y, label);
r2 = ari.measure(y, label);
System.out.format("UPGMA rand index = %.2f%%\tadjusted rand index = %.2f%%%n", 100.0 * r, 100.0 * r2);
assertTrue(r > 0.1);
hc = new HierarchicalClustering(new WPGMALinkage(proximity));
label = hc.partition(10);
r = rand.measure(y, label);
r2 = ari.measure(y, label);
System.out.format("WPGMA rand index = %.2f%%\tadjusted rand index = %.2f%%%n", 100.0 * r, 100.0 * r2);
assertTrue(r > 0.2);
hc = new HierarchicalClustering(new UPGMCLinkage(proximity));
label = hc.partition(10);
r = rand.measure(y, label);
r2 = ari.measure(y, label);
System.out.format("UPGMC rand index = %.2f%%\tadjusted rand index = %.2f%%%n", 100.0 * r, 100.0 * r2);
assertTrue(r > 0.1);
hc = new HierarchicalClustering(new WPGMCLinkage(proximity));
label = hc.partition(10);
r = rand.measure(y, label);
r2 = ari.measure(y, label);
System.out.format("WPGMC rand index = %.2f%%\tadjusted rand index = %.2f%%%n", 100.0 * r, 100.0 * r2);
assertTrue(r > 0.1);
hc = new HierarchicalClustering(new WardLinkage(proximity));
label = hc.partition(10);
r = rand.measure(y, label);
r2 = ari.measure(y, label);
System.out.format("Ward rand index = %.2f%%\tadjusted rand index = %.2f%%%n", 100.0 * r, 100.0 * r2);
assertTrue(r > 0.9);
assertTrue(r2 > 0.5);
} catch (Exception ex) {
System.err.println(ex);
}
}
use of smile.clustering.linkage.UPGMCLinkage in project smile by haifengl.
the class HierarchicalClusteringDemo method learn.
@Override
public JComponent learn() {
long clock = System.currentTimeMillis();
double[][] data = dataset[datasetIndex];
int n = data.length;
double[][] proximity = new double[n][];
for (int i = 0; i < n; i++) {
proximity[i] = new double[i + 1];
for (int j = 0; j < i; j++) proximity[i][j] = Math.distance(data[i], data[j]);
}
HierarchicalClustering hac = null;
switch(linkageBox.getSelectedIndex()) {
case 0:
hac = new HierarchicalClustering(new SingleLinkage(proximity));
break;
case 1:
hac = new HierarchicalClustering(new CompleteLinkage(proximity));
break;
case 2:
hac = new HierarchicalClustering(new UPGMALinkage(proximity));
break;
case 3:
hac = new HierarchicalClustering(new WPGMALinkage(proximity));
break;
case 4:
hac = new HierarchicalClustering(new UPGMCLinkage(proximity));
break;
case 5:
hac = new HierarchicalClustering(new WPGMCLinkage(proximity));
break;
case 6:
hac = new HierarchicalClustering(new WardLinkage(proximity));
break;
default:
throw new IllegalStateException("Unsupported Linkage");
}
System.out.format("Hierarchical clusterings %d samples in %dms\n", dataset[datasetIndex].length, System.currentTimeMillis() - clock);
int[] membership = hac.partition(clusterNumber);
int[] clusterSize = new int[clusterNumber];
for (int i = 0; i < membership.length; i++) {
clusterSize[membership[i]]++;
}
JPanel pane = new JPanel(new GridLayout(1, 3));
PlotCanvas plot = ScatterPlot.plot(dataset[datasetIndex], pointLegend);
plot.setTitle("Data");
pane.add(plot);
for (int k = 0; k < clusterNumber; k++) {
double[][] cluster = new double[clusterSize[k]][];
for (int i = 0, j = 0; i < dataset[datasetIndex].length; i++) {
if (membership[i] == k) {
cluster[j++] = dataset[datasetIndex][i];
}
}
plot.points(cluster, pointLegend, Palette.COLORS[k % Palette.COLORS.length]);
}
plot = Dendrogram.plot("Dendrogram", hac.getTree(), hac.getHeight());
plot.setTitle("Dendrogram");
pane.add(plot);
return pane;
}
Aggregations