/** Push-relabel algorithm for maximum flow */
private void push(double[][] flow, double[] excess, int u, int v) {
double send = Math.min(excess[u], graph[u][v] - flow[u][v]);
flow[u][v] += send;
flow[v][u] -= send;
excess[u] -= send;
excess[v] += send;
}
private void relabel(double[][] flow, int[] height, int u) {
int minHeight = 2 * n;
for (int v = 0; v < n; v++) {
if (graph[u][v] - flow[u][v] > 0) {
minHeight = Math.min(minHeight, height[v]);
height[u] = minHeight + 1;
}
}
}
private void _cluster(double[][] data, int k) {
long clock = System.currentTimeMillis();
SpectralClustering cluster = new SpectralClustering(data, k, 0.355);
System.out.format(
"DBSCAN clusterings %d samples in %dms\n", data.length, System.currentTimeMillis() - clock);
System.out.println("getNumClusters:" + cluster.getNumClusters());
System.out.println("getClusterSize:" + cluster.getClusterSize());
// System.out.println(JSON.toJSONString(dbscan.getClusterSize()));
System.out.println("toString:" + cluster.toString());
/** ************************************************************ */
boolean more = true;
EigenValueDecomposition eigen = cluster.getEigen();
double[] lab = eigen.getEigenValues();
double sd = smile.math.Math.sd(eigen.getEigenValues());
System.out.println("sd(eigen.getEigenValues()):" + sd);
if (Math.min(eigen.getEigenValues()) > 0.3) {
result = cluster;
cluster(data, k + 1);
} else {
return;
}
}
