Examples with SOM smile.vq.SOM used on opensource projects

Example 1 with SOM

use of smile.vq.SOM in project smile by haifengl.

the class SOMDemo method learn.

@Override
public JComponent learn() {
    try {
        width = Integer.parseInt(widthField.getText().trim());
        if (width < 1) {
            JOptionPane.showMessageDialog(this, "Invalid width: " + width, "Error", JOptionPane.ERROR_MESSAGE);
            return null;
        }
    } catch (Exception e) {
        JOptionPane.showMessageDialog(this, "Invalid width: " + widthField.getText(), "Error", JOptionPane.ERROR_MESSAGE);
        return null;
    }
    try {
        height = Integer.parseInt(heightField.getText().trim());
        if (height < 1) {
            JOptionPane.showMessageDialog(this, "Invalid height: " + height, "Error", JOptionPane.ERROR_MESSAGE);
            return null;
        }
    } catch (Exception e) {
        JOptionPane.showMessageDialog(this, "Invalid height: " + heightField.getText(), "Error", JOptionPane.ERROR_MESSAGE);
        return null;
    }
    long clock = System.currentTimeMillis();
    SOM som = new SOM(dataset[datasetIndex], width, height);
    System.out.format("SOM clusterings %d samples in %dms\n", dataset[datasetIndex].length, System.currentTimeMillis() - clock);
    JPanel pane = new JPanel(new GridLayout(2, 3));
    PlotCanvas plot = ScatterPlot.plot(dataset[datasetIndex], pointLegend);
    plot.grid(som.map());
    plot.setTitle("SOM Grid");
    pane.add(plot);
    int[] membership = som.partition(clusterNumber);
    int[] clusterSize = new int[clusterNumber];
    for (int i = 0; i < membership.length; i++) {
        clusterSize[membership[i]]++;
    }
    plot = ScatterPlot.plot(dataset[datasetIndex], pointLegend);
    plot.setTitle("Hierarchical Clustering");
    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]);
    }
    pane.add(plot);
    double[][] umatrix = som.umatrix();
    double[] umatrix1 = new double[umatrix.length * umatrix[0].length];
    for (int i = 0, k = 0; i < umatrix.length; i++) {
        for (int j = 0; j < umatrix[i].length; j++, k++) umatrix1[k] = umatrix[i][j];
    }
    plot = Histogram.plot(null, umatrix1, 20);
    plot.setTitle("U-Matrix Histogram");
    pane.add(plot);
    GaussianMixture mixture = new GaussianMixture(umatrix1);
    double w = (Math.max(umatrix1) - Math.min(umatrix1)) / 24;
    double[][] p = new double[50][2];
    for (int i = 0; i < p.length; i++) {
        p[i][0] = Math.min(umatrix1) + i * w;
        p[i][1] = mixture.p(p[i][0]) * w;
    }
    plot.line(p, Color.RED);
    plot = Hexmap.plot(umatrix, Palette.jet(256));
    plot.setTitle("U-Matrix");
    pane.add(plot);
    /*
        double[][] x = new double[height][width];
        for (int i = 0; i < height; i++) {
            for (int j = 0; j < width; j++) {
                x[i][j] = som.getMap()[i][j][0];
            }
        }
        plot = PlotCanvas.hexmap(x, Palette.jet(256));
        plot.setTitle("X");
        pane.add(plot);

        double[][] y = new double[height][width];
        for (int i = 0; i < height; i++) {
            for (int j = 0; j < width; j++) {
                y[i][j] = som.getMap()[i][j][1];
            }
        }
        plot = PlotCanvas.hexmap(y, Palette.jet(256));
        plot.setTitle("Y");
        pane.add(plot);
*/
    return pane;
}

Example 2 with SOM

use of smile.vq.SOM in project smile by haifengl.

the class HexmapDemo method main.

public static void main(String[] args) {
    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 m = 20;
        int n = 20;
        SOM som = new SOM(x, m, n);
        String[][] labels = new String[m][n];
        int[] neurons = new int[x.length];
        for (int i = 0; i < x.length; i++) {
            neurons[i] = som.predict(x[i]);
        }
        int[] count = new int[10];
        for (int i = 0; i < m; i++) {
            for (int j = 0; j < n; j++) {
                Arrays.fill(count, 0);
                for (int k = 0; k < neurons.length; k++) {
                    if (neurons[k] == i * n + j) {
                        count[y[k]]++;
                    }
                }
                int sum = Math.sum(count);
                if (sum == 0.0) {
                    labels[i][j] = "no samples";
                } else {
                    labels[i][j] = String.format("<table border=\"1\"><tr><td>Total</td><td align=\"right\">%d</td></tr>", sum);
                    for (int l = 0; l < count.length; l++) {
                        if (count[l] > 0) {
                            labels[i][j] += String.format("<tr><td>class %d</td><td align=\"right\">%.1f%%</td></tr>", l, 100.0 * count[l] / sum);
                        }
                    }
                    labels[i][j] += "</table>";
                }
            }
        }
        double[][] umatrix = som.umatrix();
        double[][][] map = som.map();
        double[][] proximity = new double[m * n][m * n];
        for (int i = 0; i < m * n; i++) {
            for (int j = 0; j < m * n; j++) {
                proximity[i][j] = Math.distance(map[i / n][i % n], map[j / n][j % n]);
            }
        }
        MDS mds = new MDS(proximity, 3);
        double[][] coords = mds.getCoordinates();
        double[][][] mdsgrid = new double[m][n][];
        for (int i = 0; i < m * n; i++) {
            mdsgrid[i / n][i % n] = mds.getCoordinates()[i];
        }
        SammonMapping sammon = new SammonMapping(proximity, coords);
        double[][][] sammongrid = new double[m][n][];
        for (int i = 0; i < m * n; i++) {
            sammongrid[i / n][i % n] = sammon.getCoordinates()[i];
        }
        IsotonicMDS isomds = new IsotonicMDS(proximity, coords);
        double[][][] isomdsgrid = new double[m][n][];
        for (int i = 0; i < m * n; i++) {
            isomdsgrid[i / n][i % n] = isomds.getCoordinates()[i];
        }
        JFrame frame = new JFrame("Hexmap");
        frame.setDefaultCloseOperation(JFrame.EXIT_ON_CLOSE);
        frame.setLocationRelativeTo(null);
        frame.add(Hexmap.plot(labels, umatrix));
        PlotCanvas canvas = Surface.plot(mdsgrid);
        canvas.setTitle("MDS");
        frame.add(canvas);
        canvas = Surface.plot(isomdsgrid);
        canvas.setTitle("Isotonic MDS");
        frame.add(canvas);
        canvas = Surface.plot(sammongrid);
        canvas.setTitle("Sammon Mapping");
        frame.add(canvas);
        frame.setVisible(true);
    } catch (Exception ex) {
        System.err.println(ex);
    }
}