Examples with Clusters edu.ucsf.rbvi.clusterMaker2.internal.algorithms.attributeClusterers.Clusters used on opensource projects

Example 36 with Clusters

use of edu.ucsf.rbvi.clusterMaker2.internal.algorithms.attributeClusterers.Clusters in project clusterMaker2 by RBVI.

the class RunFFT method kcluster.

@Override
public int kcluster(int nClusters, int nIterations, CyMatrix matrix, DistanceMetric metric, int[] clusterID) {
    random = null;
    int nelements = matrix.nRows();
    int[] tclusterid = new int[nelements];
    int[] saved = new int[nelements];
    int[] mapping = new int[nClusters];
    int[] counts = new int[nClusters];
    CyMatrix distanceMatrix = matrix.getDistanceMatrix(metric);
    HashMap<Integer, Integer> centers = new HashMap<Integer, Integer>();
    double error = Double.MAX_VALUE;
    if (monitor != null)
        monitor.setProgress(0);
    for (int i = 0; i < nelements; i++) clusterID[i] = 0;
    // the first center
    Random randomGenerator = new Random();
    centers.put(0, randomGenerator.nextInt(nelements));
    // now find the remaining centers
    for (int i = 1; i < nClusters; i++) {
        int y = getMaxMin(centers, distanceMatrix);
        centers.put(i, y);
        clusterID[y] = i;
    }
    /*
		for (int i = 0; i < nClusters; i++)
			System.out.printf("Center for %d = %d\n", i, centers.get(i));
		*/
    // assign clusters now
    int k = centers.get(0);
    for (int i = 0; i < nelements; i++) {
        // Is this one of our centers?
        if (centers.containsValue(i))
            continue;
        double minDistance = Double.MAX_VALUE;
        int center = 0;
        for (int cluster = 0; cluster < nClusters; cluster++) {
            double dist = distanceMatrix.doubleValue(i, centers.get(cluster));
            if (dist < minDistance) {
                center = cluster;
                minDistance = dist;
            }
        }
        clusterID[i] = center;
    }
    return nClusters;
}

Example 37 with Clusters

use of edu.ucsf.rbvi.clusterMaker2.internal.algorithms.attributeClusterers.Clusters in project clusterMaker2 by RBVI.

the class RunAutoSOME method run.

public List<NodeCluster> run(TaskMonitor monitor) {
    long startTime = System.currentTimeMillis();
    debugln("Initial matrix:");
    // printMatrix(matrix);
    // Normalize
    // normalize(matrix, clusteringThresh, false);
    debugln("Normalized matrix:");
    if (dataAttributes == null || dataAttributes.isEmpty()) {
        monitor.showMessage(TaskMonitor.Level.ERROR, "Error: no attribute list selected");
        return null;
    }
    if (selectedOnly && network.getDefaultNodeTable().countMatchingRows(CyNetwork.SELECTED, true) == 0) {
        monitor.showMessage(TaskMonitor.Level.ERROR, "Error: no nodes selected from network");
        return null;
    }
    Settings s = new Settings();
    // get parameter settings
    s.ensemble_runs = settings.ensemble_runs;
    s.mst_pval = settings.mst_pval;
    s.threads = settings.threads;
    s.logNorm = settings.logNorm;
    s.unitVar = settings.unitVar;
    s.medCenter = settings.medCenter;
    s.medCenterCol = settings.medCenterCol;
    s.sumSqrRows = settings.sumSqrRows;
    s.sumSqrCol = settings.sumSqrCol;
    s.som_iters = settings.som_iters;
    s.de_resolution = settings.de_resolution;
    s.distMatrix = settings.distMatrix;
    s.dmDist = settings.dmDist;
    s.FCNrows = settings.FCNrows;
    s.som_maxGrid = 20;
    s.htmlOut = false;
    s.textOut = false;
    // System.out.println("network "+network.toString()+" has "+network.getNodeCount()+" nodes");
    // System.out.println("attributes: "+dataAttributes);
    String[] attrArray = new String[dataAttributes.size()];
    int att = 0;
    for (String attribute : dataAttributes) {
        attrArray[att++] = "node." + attribute;
    }
    // Create the matrix
    CyMatrix matrix = CyMatrixFactory.makeSmallMatrix(network, attrArray, selectedOnly, ignoreMissing, false, false);
    if (!selectedOnly) {
        nodes = network.getNodeList();
    } else {
        nodes = CyTableUtil.getNodesInState(network, CyNetwork.SELECTED, true);
    }
    s.input = new dataItem[matrix.nRows()];
    // matrix.printMatrix();
    Map<String, Integer> key = new HashMap<String, Integer>();
    for (int i = 0; i < nodes.size(); i++) {
        String id = ModelUtils.getNodeName(network, nodes.get(i));
        if (!key.containsKey(id))
            key.put(id, i);
    }
    for (int k = 0, itor = 0; k < matrix.nRows(); k++) {
        float[] f = new float[matrix.nColumns()];
        // System.out.println(matrix.getRowLabels()[k]+" "+nodes.get(k).getIdentifier());
        if (k == 0) {
            s.columnHeaders = new String[f.length + 1];
            s.columnHeaders[0] = "NAME";
        }
        for (int l = 0; l < f.length; l++) {
            if (k == 0) {
                s.columnHeaders[l + 1] = matrix.getColumnLabel(l);
                s.columnHeaders[l + 1] = s.columnHeaders[l + 1].replace("\"", "");
                s.columnHeaders[l + 1] = s.columnHeaders[l + 1].replace(",", "");
            // System.out.println(s.columnHeaders[l+1]);
            }
            // System.out.println(matrix.getValue(k,l).floatValue());
            if (matrix.getValue(k, l) != null) {
                f[l] = matrix.getValue(k, l).floatValue();
            } else {
                f[l] = -99999999;
                s.fillMissing = true;
            }
        }
        s.input[itor++] = new dataItem(f, matrix.getRowLabel(k));
    }
    if (s.FCNrows && s.distMatrix)
        s = transpose(s);
    if (s.input == null) {
        monitor.showMessage(TaskMonitor.Level.ERROR, "Insufficient data for clustering (1 or less rows or columns)");
        return null;
    } else if (s.input.length < 2) {
        monitor.showMessage(TaskMonitor.Level.ERROR, "Insufficient data for clustering (1 or less rows or columns)");
        return null;
    } else if (s.input[0].getValues() == null) {
        monitor.showMessage(TaskMonitor.Level.ERROR, "Insufficient data for clustering (1 or less rows or columns)");
        return null;
    } else if (s.input[0].getValues().length < 2) {
        monitor.showMessage(TaskMonitor.Level.ERROR, "Insufficient data for clustering (1 or less rows or columns)");
        return null;
    }
    autRun = new Run();
    cr = autRun.runAutoSOMEBasic(s, monitor);
    if (cr == null) {
        monitor.setStatusMessage("Clustering failed!");
        return null;
    }
    monitor.setStatusMessage("Assigning nodes to clusters");
    clusterCount = cr.c.length;
    Map<NodeCluster, NodeCluster> cMap = (!s.distMatrix) ? getNodeClusters(cr, key, matrix, s) : getNodeClustersFCN(cr, matrix, s);
    if (canceled) {
        monitor.setStatusMessage("canceled");
        return null;
    }
    // Update node attributes in network to include clusters. Create cygroups from clustered nodes
    monitor.setStatusMessage("Created " + clusterCount + " clusters");
    // 
    if (clusterCount == 0) {
        monitor.showMessage(TaskMonitor.Level.WARN, "Created 0 clusters!!!!");
        return null;
    }
    Set<NodeCluster> clusters = cMap.keySet();
    return new ArrayList<NodeCluster>(clusters);
}

Example 38 with Clusters

use of edu.ucsf.rbvi.clusterMaker2.internal.algorithms.attributeClusterers.Clusters in project clusterMaker2 by RBVI.

the class PP_DivideAndReclusterRecursively method recursiveReclustering.

/**
 * This method goes through each cluster in the tmpClusterObject and
 * re-clusters them. If the clusters have less than 4 objects, then it is
 * added to the finalClusterObject. If after the re-clustering, no changes
 * have occured (it is still one cluster) it is added to the
 * finalClusterObject. Otherwise if changes have occured the sub-clusters
 * are then re-clustered again. This occurs recursively until no more
 * improvements are found.
 *
 * @param finalClusterObject
 *            This should only include finished clusters that can not be
 *            improved anymore by reclustering.
 * @param tmpClusterObject
 *            This includes all clusters that are to be re-clustered.
 */
private void recursiveReclustering(ArrayList<ArrayList<Integer>> finalClusterObject, ArrayList<ArrayList<Integer>> tmpClusterObject) {
    for (int i = 0; i < tmpClusterObject.size(); i++) {
        ArrayList<Integer> cluster = tmpClusterObject.get(i);
        int clusterSize = cluster.size();
        /*all clusters <=3 are taken care of in the rearrange method of
			 * post-processing */
        if (clusterSize <= 3) {
            finalClusterObject.add(cluster);
        } else {
            /* run clustering for this one cluster to see if an improvement can be found */
            ConnectedComponent ccForCluster = this.cc.createConnectedComponentForCluster(i, cluster);
            ClusterPostProcessingTask clusterTask = new ClusterPostProcessingTask(ccForCluster, this.params, this.layouterEnumTypes);
            clusterTask.run();
            /* if there has been no change, add this cluster to the final cluster object */
            if (ccForCluster.getNumberOfClusters() == 1) {
                finalClusterObject.add(cluster);
            /* otherwise recluster */
            } else {
                /* new cluster object for the resulting clusters of the improved cluster */
                ArrayList<ArrayList<Integer>> nextClusterObject = PostProcessingUtility.createClusterObject(ccForCluster, true);
                recursiveReclustering(finalClusterObject, nextClusterObject);
            }
        }
    }
}

Example 39 with Clusters

use of edu.ucsf.rbvi.clusterMaker2.internal.algorithms.attributeClusterers.Clusters in project clusterMaker2 by RBVI.

the class ClusteringManager method initParametersAndCCs.

/**
 * Sets transitive connected components file if in input directory. Finds all cost matrix files
 * given and creates a new {@link ConnectedComponent} instance for each and adds them to
 * the list of ConnectedComponents.
 *
 * Also initialises the {@link IParameters} from the input configuration.
 * @throws InvalidInputFileException
 */
public void initParametersAndCCs() throws InvalidInputFileException {
    if (TaskConfig.mode == TaskConfig.CLUSTERING_MODE || TaskConfig.mode == TaskConfig.GENERAL_TRAINING_MODE) {
        // this.connectedComponents = new ArrayList<ConnectedComponent>();
        this.connectedComponents = new ArrayList<File>();
        TaskConfig.transitiveConnectedComponents = null;
        /* read the input file or directory */
        File cmFile = new File(this.cmPath);
        if (cmFile.isDirectory()) {
            log.finer("Input is a directory!");
            /* get all files in the directory */
            File[] files = cmFile.listFiles();
            /* check boolean whether cm files exist in directory */
            boolean noCostMatrices = true;
            boolean noTCCfile = true;
            for (int i = 0; i < files.length; i++) {
                String filePath = files[i].toString();
                /* find tcc file in directory */
                if (filePath.endsWith(".tcc") || filePath.endsWith(".rtcc")) {
                    noTCCfile = false;
                    TaskConfig.transitiveConnectedComponents = filePath;
                    log.info("Transitive connected components file: " + filePath);
                    InfoFile.appendToProjectDetails("Transitive connected component file: " + filePath);
                }
                /* find cm files*/
                if (files[i].toString().endsWith(".cm") || files[i].toString().endsWith(".rcm")) {
                    // cm files exist
                    noCostMatrices = false;
                    connectedComponents.add(files[i]);
                // create the connected component (cc) object and add to list
                // CostMatrixReader cmReader = new CostMatrixReader(files[i]);
                // ConnectedComponent cc = cmReader.getConnectedComponent();
                // connectedComponents.add(cc);
                }
            }
            if (noCostMatrices) {
                if (noTCCfile) {
                    throw new InvalidInputFileException("There are no cost matrix " + "files in the input directory and also no transitive connected components file, " + "or check whether the file extensions equal .cm, .rcm, or .tcc");
                }
            }
        } else {
            /* only one cost matrix file is given */
            log.finer("One cm file given");
            /* only one cost matrix as input - start clustering process */
            if (!cmFile.toString().endsWith(".tcc")) {
                // create the connected component (cc) object and add to list
                // CostMatrixReader cmReader = new CostMatrixReader(cmFile);
                // ConnectedComponent cc = cmReader.getConnectedComponent();
                // connectedComponents.add(cc);
                connectedComponents.add(cmFile);
            } else {
                if (cmFile.toString().endsWith(".tcc")) {
                    TaskConfig.transitiveConnectedComponents = cmFile.toString();
                    log.info("Only a transitive connected component file is given: " + cmFile.toString());
                    InfoFile.appendToProjectDetails("Only a transitive connected component file is given: " + cmFile.toString() + ". Therefore NO CLUSTERING IS PERFORMED, just the the " + "clusters from the TCC file are written into the clusters file.");
                } else {
                    throw new InvalidInputFileException("Either the input cost matrix is of " + "wrong file type. The file extension should be \".cm\" or \".rcm\"," + "or in the given TCC file is of the wrong type and should be \".tcc\".");
                }
            }
        }
    }
    /* initialise parameters from config */
    // LayoutFactory.EnumLayouterClass[] layouterEnumTypes = TaskConfig.layouterEnumTypes;
    layouterParameters = new IParameters[TaskConfig.layouterEnumTypes.length];
    for (int i = 0; i < TaskConfig.layouterEnumTypes.length; i++) {
        IParameters param = TaskConfig.layouterEnumTypes[i].createIParameters();
        param.readParametersFromConfig();
        layouterParameters[i] = param;
    }
    if (this.connectedComponents == null) {
    }
}

Example 40 with Clusters

use of edu.ucsf.rbvi.clusterMaker2.internal.algorithms.attributeClusterers.Clusters in project clusterMaker2 by RBVI.

the class ClusteringManager method runClusteringForOneConnectedComponent.

/**
 * Runs clustering for one {@link ConnectedComponent} and sets the total score to
 * the parameters if in the general training mode.
 * @param cc The connected component object.
 * @param clusterFile The clusters file (null if in general training mode)
 * @param semaphore The Semaphore to give to the clustering task to keep track of it.
 * @param time
 * @throws InvalidInputFileException
 */
public void runClusteringForOneConnectedComponent(ConnectedComponent cc, ClusterFile clusterFile, Semaphore semaphore, Semaphore maxThreadSemaphore, long time) throws InvalidInputFileException {
    /* check whether layouterParameters has been initialised */
    if (this.layouterParameters == null) {
        if (TaskConfig.mode == TaskConfig.CLUSTERING_MODE || TaskConfig.mode == TaskConfig.GENERAL_TRAINING_MODE)
            log.warning("Incorrect use of the ClusteringManager, the layouter parameters list" + "hadn't been initialised. Called method to initialise this and the connected components from " + "the config");
        this.initParametersAndCCs();
    }
    // if(TaskConfig.mode == TaskConfig.CLUSTERING_MODE){
    if (TaskConfig.doLayoutParameterTraining && !TaskConfig.greedy) {
        for (int i = 0; i < this.layouterParameters.length; i++) {
            /* start parameter training for the cc */
            IParameterTraining paramTrain = TaskConfig.parameterTrainingEnum.createParameterTrainer();
            paramTrain.initialise(TaskConfig.layouterEnumTypes[i], TaskConfig.noOfParameterConfigurationsPerGeneration, TaskConfig.noOfGenerations);
            paramTrain.setMaxThreadSemaphoreAndThreadsList(maxThreadSemaphore, this.allThreads);
            IParameters bestparam = paramTrain.run(cc);
            log.fine("PARAMETER TRAINING RESULT\n: " + cc.getCcPath() + "\n" + bestparam.toString());
            this.layouterParameters[i] = bestparam;
        }
    }
    // }
    /* run clustering with the previously determined parameters */
    ClusteringTask clusterTask = new ClusteringTask(cc, this.layouterParameters, TaskConfig.layouterEnumTypes, clusterFile);
    clusterTask.setTime(time);
    // if(!TaskConfig.doLayoutParameterTraining&&TaskConfig.useThreads){
    // clusterTask.setSemaphore(semaphore);
    // Thread t = new Thread(clusterTask);
    // clusterTask.setMaxThreadSemaphore(maxThreadSemaphore, allThreads, t);
    // t.start();
    // }else{
    clusterTask.run();
// }
}