/**
* Write the clusterings out into a file.
*
* @param fileName the file to write out to
* @param clustering the clusters to write out
* @throws IOException occurs if there's an input or output error
*/
private static void writeOutputFile(
String fileName, Clustering clustering, boolean adjustedEdgeWeights) throws IOException {
BufferedWriter bufferedWriter;
int i, nNodes;
// get the information
nNodes = clustering.getNNodes();
clustering.orderClustersByNNodes();
// writing to the .graph: final communities output
if (adjustedEdgeWeights) {
fileName = fileName + "-lm-performance" + "-rehoM" + ".graph";
} else {
fileName = fileName + "-lm-performance" + "-maxM" + ".graph";
}
bufferedWriter = new BufferedWriter(new FileWriter(fileName));
for (i = 0; i < nNodes; i++) {
bufferedWriter.write(i + " " + Integer.toString(clustering.getCluster(i)));
bufferedWriter.newLine();
}
bufferedWriter.close();
}
/**
* Resets the values of clusters that have been merged.
*
* @param clustering
* @param i
* @param j
*/
protected void updateScoreMatrix(Clustering clustering, int i, int j) {
int size = clustering.getNumInstances();
int[] ci = clustering.getIndicesWithLabel(i);
for (int ni = 0; ni < ci.length; ni++) {
for (int nj = 0; nj < size; nj++) if (ci[ni] != nj) scoreCache.set(ci[ni], nj, 0.0);
}
int[] cj = clustering.getIndicesWithLabel(j);
for (int ni = 0; ni < cj.length; ni++) {
for (int nj = 0; nj < size; nj++) if (cj[ni] != nj) scoreCache.set(cj[ni], nj, 0.0);
}
}
/**
* @param clustering
* @param i
* @param j
* @return The score for merging these two clusters.
*/
protected double getScore(Clustering clustering, int i, int j) {
if (scoreCache == null) scoreCache = new PairwiseMatrix(clustering.getNumInstances());
int[] ci = clustering.getIndicesWithLabel(i);
int[] cj = clustering.getIndicesWithLabel(j);
if (scoreCache.get(ci[0], cj[0]) == 0.0) {
double val =
evaluator.evaluate(
new AgglomerativeNeighbor(
clustering, ClusterUtils.copyAndMergeClusters(clustering, i, j), ci, cj));
for (int ni = 0; ni < ci.length; ni++)
for (int nj = 0; nj < cj.length; nj++) scoreCache.set(ci[ni], cj[nj], val);
}
return scoreCache.get(ci[0], cj[0]);
}
public static void testLoadingData(String[] args) {
System.out.println("Loading point cloud from data: ex 0");
if (args.length < 1) {
System.out.println("Usage: java ImageManipulation filename.dat");
System.exit(0);
}
// load point cloud
PointCloud N;
// N=PointCloud.randomPoints(10000, 3);
N = Clustering.readFile(args[0]);
Draw.draw2D(N, "original point cloud");
Draw.draw3D(N);
}
/**
* For each pair of clusters, calculate the score of the {@link
* cc.mallet.cluster.neighbor_evaluator.Neighbor} that would result from merging the two clusters.
* Choose the merge that obtains the highest score. If no merge improves score, return original
* Clustering
*
* @param clustering
* @return
*/
public Clustering improveClustering(Clustering clustering) {
double bestScore = Double.NEGATIVE_INFINITY;
int[] toMerge = new int[] {-1, -1};
for (int i = 0; i < clustering.getNumClusters(); i++) {
for (int j = i + 1; j < clustering.getNumClusters(); j++) {
double score = getScore(clustering, i, j);
if (score > bestScore) {
bestScore = score;
toMerge[0] = i;
toMerge[1] = j;
}
}
}
converged = (bestScore < stoppingThreshold);
if (!(converged)) {
progressLogger.info(
"Merging "
+ toMerge[0]
+ "("
+ clustering.size(toMerge[0])
+ " nodes) and "
+ toMerge[1]
+ "("
+ clustering.size(toMerge[1])
+ " nodes) ["
+ bestScore
+ "] numClusters="
+ clustering.getNumClusters());
updateScoreMatrix(clustering, toMerge[0], toMerge[1]);
clustering = ClusterUtils.mergeClusters(clustering, toMerge[0], toMerge[1]);
} else {
progressLogger.info("Converged with score " + bestScore);
}
return clustering;
}
/**
* The main sequence. TODO
*
* @param args expects 9 arguments
* @throws IOException occurs if there's an input or output error
*/
public static void main(String[] args) throws IOException {
boolean printOutput, update, adjustedEdgeWeights = false;
Clustering clustering;
Console console;
double performance, maxPerformance, resolution, resolution2, meaningfulMaxM, v_scalingParam;
int i, j, nIterations, nRandomStarts;
long beginTime, endTime, randomSeed;
Network network;
Random random;
String inputFileName, outputFileName;
VOSClusteringTechnique VOSClusteringTechnique;
System.out.println("-------------------------------------");
System.out.println("Modularity Optimizer");
System.out.println("Version 1.3.0");
System.out.println("by Ludo Waltman and Nees Jan van Eck");
System.out.println("-------------------------------------");
// read in the arguments
if (args.length == 9) {
inputFileName = args[0];
outputFileName = args[1];
v_scalingParam = Double.parseDouble(args[2]);
resolution = Double.parseDouble(args[3]);
meaningfulMaxM = Double.parseDouble(args[4]);
nRandomStarts = Integer.parseInt(args[5]);
nIterations = Integer.parseInt(args[6]);
randomSeed = Long.parseLong(args[7]);
printOutput = (Integer.parseInt(args[8]) > 0);
} else {
console = System.console();
inputFileName = console.readLine("Input file path/name: ");
outputFileName = console.readLine("Output file path/name (without file extension): ");
v_scalingParam =
Double.parseDouble(
console.readLine(
"Scaling parameter V (0 to 1; rates the"
+ " importance of the weight of inter-cluster edges,"
+ " with respect to the weight of intra-cluster edges): "));
resolution = Double.parseDouble(console.readLine("Resolution parameter (e.g., 1.0): "));
meaningfulMaxM =
Double.parseDouble(console.readLine("Meaningful maximum M of edge weights: "));
nRandomStarts = Integer.parseInt(console.readLine("Number of random starts (e.g., 10): "));
nIterations = Integer.parseInt(console.readLine("Number of iterations (e.g., 10): "));
randomSeed = Long.parseLong(console.readLine("Random seed (e.g., 0): "));
printOutput = (Integer.parseInt(console.readLine("Print output (0 = no; 1 = yes): ")) > 0);
System.out.println();
}
// read input file
if (printOutput) System.out.println("Reading input file...");
network = readInputFile(inputFileName);
if (printOutput) System.out.println("Finish reading input file.");
// print network characteristics
if (printOutput) {
System.out.println();
System.out.format("Number of nodes: %d%n", network.getNNodes());
System.out.format("Number of edges: %d%n", network.getNEdges());
System.out.println("Running Louvain algorithm - performance");
System.out.println("Scaling parameter V: " + v_scalingParam);
System.out.println("Meaningful maximum M: " + meaningfulMaxM);
System.out.println();
}
/**
* Run the Louvain algorithm nRandomStarts times, beginning with a random node configuration
* every time. After all the trials, return the trial that resulted in the best performance
* score.
*/
beginTime = System.currentTimeMillis();
clustering = null;
maxPerformance = Double.NEGATIVE_INFINITY;
random = new Random(randomSeed);
for (i = 0; i < nRandomStarts; i++) {
if (printOutput && (nRandomStarts > 1)) {
System.out.format("\tRandom start: %d%n", i + 1);
}
VOSClusteringTechnique =
new VOSClusteringTechnique(
v_scalingParam, meaningfulMaxM, network, resolution, outputFileName);
j = 0;
update = true;
// Print performance of unaltered graph once
if (i == 0) {
System.out.println(
"Performance of unaltered graph: " + VOSClusteringTechnique.calcPerformanceFunction());
}
// run for nIterations times, or until there is no increase in performance
do {
if (printOutput && (nIterations > 0)) {
System.out.format("\nIteration: %d%n", j + 1);
update = VOSClusteringTechnique.runLouvainAlgorithm(random);
}
j++;
performance = VOSClusteringTechnique.calcPerformanceFunction();
if (printOutput && (nIterations > 0))
System.out.format("\nIteration %d performance: %.4f%n", j, performance);
} while ((j < nIterations) && update);
// we have found our best clustering
if (performance > maxPerformance) {
clustering = VOSClusteringTechnique.getClustering();
maxPerformance = performance;
}
/* Print all clusterings to .tree file */
VOSClusteringTechnique.printClusteringsToTree();
/* Find the best clustering and print it */
VOSClusteringTechnique.bestOverallClustering();
/* Are we in the M = max or M = remove high outliers case? */
adjustedEdgeWeights = VOSClusteringTechnique.adjustedEdgeWeights;
}
endTime = System.currentTimeMillis();
// print end-of-algorithm information
if (printOutput) {
System.out.format(
"\nMaximum performance in %d random starts: %.4f%n", nRandomStarts, maxPerformance);
System.out.format("Number of communities: %d%n", clustering.getNClusters());
System.out.format("Elapsed time: %d seconds%n", Math.round((endTime - beginTime) / 1000.0));
System.out.println();
}
// write to output file
if (printOutput) System.out.println("Writing to non-optimized .graph...");
writeOutputFile(outputFileName, clustering, adjustedEdgeWeights);
if (printOutput) System.out.println("Finish writing to non-optimized .graph.");
}
