public XChartPanel buildPanel() throws IOException {
System.out.println("fetching data...");
updateData();
// create chart
Chart chart =
new ChartBuilder()
.chartType(ChartType.Area)
.width(800)
.height(400)
.title("Real-time Bitcoinium Order Book - BITSTAMP_BTC_USD")
.xAxisTitle("BTC")
.yAxisTitle("USD")
.build();
chart.getStyleManager().setLegendPosition(LegendPosition.InsideNE);
// add series
Series series = chart.addSeries(BIDS_SERIES_NAME, xAxisBidData, yAxisBidData);
series.setMarker(SeriesMarker.NONE);
series = chart.addSeries(ASKS_SERIES_NAME, xAxisAskData, yAxisAskData);
series.setMarker(SeriesMarker.NONE);
return new XChartPanel(chart);
}
/**
* Makes a plot showing overlaid individual normalized count for every word in WORDS from
* STARTYEAR to ENDYEAR using NGM as a data source.
*/
public static void plotAllWords(NGramMap ngm, String[] words, int startYear, int endYear) {
Chart chart =
new ChartBuilder().width(800).height(600).xAxisTitle("years").yAxisTitle("data").build();
for (String word : words) {
TimeSeries bundle = ngm.weightHistory(word, startYear, endYear);
chart.addSeries(word, bundle.years(), bundle.data());
}
new SwingWrapper(chart).displayChart();
}
/**
* Creates overlaid category weight plots for each category label in CATEGORYLABELS from STARTYEAR
* to ENDYEAR using NGM and WN as data sources.
*/
public static void plotCategoryWeights(
NGramMap ngm, WordNet wn, String[] categoryLabels, int startYear, int endYear) {
Chart chart =
new ChartBuilder().width(800).height(600).xAxisTitle("years").yAxisTitle("data").build();
for (String categoryLabel : categoryLabels) {
Set words = wn.hyponyms(categoryLabel);
TimeSeries bundle = ngm.summedWeightHistory(words, startYear, endYear);
chart.addSeries(categoryLabel, bundle.years(), bundle.data());
}
new SwingWrapper(chart).displayChart();
}
/**
* Plots the normalized count of every word against the rank of every word on a log-log plot. Uses
* data from YEAR, using NGM as a data source.
*/
public static void plotZipfsLaw(NGramMap ngm, int year) {
YearlyRecord yr = ngm.getRecord(year);
Collection counts = yr.counts();
Collection ranks = downRange(counts.size());
Chart chart =
new ChartBuilder().width(800).height(600).xAxisTitle("rank").yAxisTitle("count").build();
chart.getStyleManager().setYAxisLogarithmic(true);
chart.getStyleManager().setXAxisLogarithmic(true);
chart.addSeries("zipf", ranks, counts);
new SwingWrapper(chart).displayChart();
}
@Override
public Chart getChart() {
yData = getRandomData(5);
// Create Chart
Chart chart = new Chart(500, 400);
chart.setChartTitle("Sample Real-time Chart");
chart.setXAxisTitle("X");
chart.setYAxisTitle("Y");
chart.addSeries(SERIES_NAME, null, yData);
return chart;
}
@Override
public Chart getChart() {
// Create Chart
Chart chart = new Chart(800, 600);
// Customize Chart
chart.setChartTitle("LineChart05");
chart.setXAxisTitle("X");
chart.setYAxisTitle("Y");
chart.getStyleManager().setLegendPosition(LegendPosition.InsideSW);
double[] xData = new double[] {0.0, 1.0, 2.0, 3.0, 4.0, 5, 6};
double[] yData = new double[] {106, 44, 26, 10, 7.5, 3.4, .88};
double[] yData2 = new double[] {102, 49, 23.6, 11.3, 5.4, 2.6, 1.25};
Series series = chart.addSeries("A", xData, yData);
series.setLineStyle(SeriesLineStyle.NONE);
series.setMarker(SeriesMarker.DIAMOND);
series.setMarkerColor(Color.BLACK);
Series series2 = chart.addSeries("B", xData, yData2);
series2.setMarker(SeriesMarker.NONE);
series2.setLineStyle(SeriesLineStyle.DASH_DASH);
series2.setLineColor(Color.BLACK);
chart.getStyleManager().setYAxisLogarithmic(true);
chart.getStyleManager().setYAxisMin(0.01);
chart.getStyleManager().setYAxisMax(1000);
chart.getStyleManager().setXAxisMin(2);
chart.getStyleManager().setXAxisMax(7);
return chart;
}
private void go() throws IOException {
// initialize the ahahNode
double readPeriod = 1E-5; // increase the learning rate from default
double writePeriod = 1E-5;
int numBias = 15;
int numInputs = 70; // max number of inputs/spikes
MemristorType memristorType = MemristorType.AgChalc;
AHaH21Circuit ahahNodeCircuit =
new AHaH21CircuitBuilder()
.numInputs(numInputs)
.numBiasInputs(numBias)
.readPeriod(readPeriod)
.writePeriod(writePeriod)
.memristorType(memristorType)
.build();
// initialize the feature factory. This code converts the raw data to a spiking representation.
BreastCancerSpikeEncoder breastCancerSpikeEncoder = new BreastCancerSpikeEncoder();
SpikeConverter spikeConverter =
new SpikeConverter(); // converts the spike code produced by the feature factory to one that
// can be used by the circuit.
// train the classifier--->
List<BreastCancer> breastCancerTrainingData = BreastCancerDAO.selectTrainData();
System.out.println("Learning...");
for (BreastCancer breastCancer : breastCancerTrainingData) {
String[] topicTrueLabels = new String[1];
topicTrueLabels[0] = breastCancer.getCellClass() + "";
long[] spikes = breastCancerSpikeEncoder.encode(breastCancer); // get the spike representation
Set<Integer> convertedSpikes = spikeConverter.convert(spikes);
int superviseSignal = 0;
if (breastCancer.getCellClass() == 2) {
superviseSignal = 1;
} else {
superviseSignal = -1;
}
System.out.println("Spike Pattern Length = " + convertedSpikes.size());
// System.out.println("Spikes = " + Arrays.toString(convertedSpikes.toArray()));
double y = ahahNodeCircuit.update(convertedSpikes, superviseSignal);
}
System.out.println("Spike Pattern Space = " + breastCancerSpikeEncoder.getSpikePatternSpace());
// Test the classifier
List<BreastCancer> breastCancerTestData = BreastCancerDAO.selectTestData();
System.out.println("Testing...");
Set<String> labels =
new HashSet<String>(); // these are the labels we are evaluating the performance of
labels.add("2");
labels.add("4");
// we are testing the performance for a range of confidence values, from 0 to 1.
int numSteps = 100;
double[] confidenceThresholds = new double[numSteps];
double inc = .061 / numSteps;
for (int i = 0; i < confidenceThresholds.length; i++) {
confidenceThresholds[i] = i * inc;
}
// create evaluators for keeping track of performance
ClassificationEvaluator[] evaluators = new ClassificationEvaluator[numSteps];
for (int i = 0; i < evaluators.length; i++) {
evaluators[i] = new ClassificationEvaluator(labels);
}
// run through the test set
for (BreastCancer breastCancer : breastCancerTestData) {
long[] spikes = breastCancerSpikeEncoder.encode(breastCancer);
Set<Integer> convertedSpikes = spikeConverter.convert(spikes);
double y =
ahahNodeCircuit.update(convertedSpikes, 0); // NOTE: no supervised labels are passed here.
// true labels for evaluation.
Set<String> trueLabels = new HashSet<String>();
trueLabels.add(breastCancer.getCellClass() + "");
for (int i = 0; i < evaluators.length; i++) {
List<String> labelOutput = new ArrayList<String>();
if (y > confidenceThresholds[i]) {
labelOutput.add("2");
} else if (y < -confidenceThresholds[i]) {
labelOutput.add("4");
}
evaluators[i].update(trueLabels, labelOutput);
}
}
// plot results---->
double maxF1 = 0;
int cIdx = 0;
double[] accuracy = new double[numSteps];
double[] precision = new double[numSteps];
double[] recall = new double[numSteps];
double[] f1 = new double[numSteps];
for (int i = 0; i < numSteps; i++) {
accuracy[i] = evaluators[i].getAccuracyMicroAve();
precision[i] = evaluators[i].getPrecisionMicroAve();
recall[i] = evaluators[i].getRecallMicroAve();
f1[i] = evaluators[i].getF1MicroAve();
if (f1[i] > maxF1) { // get the best confidence threshold
maxF1 = f1[i];
cIdx = i;
}
}
Chart chart = new Chart(300, 300, ChartTheme.Matlab);
// chart.setChartTitle("Wisconsin Breast Cancer Benchmark - F1=" + df.format(maxF1) + "
// w/ConfidenceThreshold=" + df.format(confidenceThresholds[cIdx]));
chart.setChartTitle("Breast Cancer - Circuit");
chart.setXAxisTitle("Confidence Threshold");
chart.setYAxisTitle("Score");
chart.getStyleManager().setLegendPosition(LegendPosition.InsideSW);
Series accuracy_series = chart.addSeries("Accuracy", confidenceThresholds, accuracy);
accuracy_series.setMarker(SeriesMarker.NONE);
Series precision_series = chart.addSeries("Precision", confidenceThresholds, precision);
precision_series.setMarker(SeriesMarker.NONE);
Series recall_series = chart.addSeries("Recall", confidenceThresholds, recall);
recall_series.setMarker(SeriesMarker.NONE);
Series f1_series = chart.addSeries("F1", confidenceThresholds, f1);
f1_series.setMarker(SeriesMarker.NONE);
BitmapEncoder.savePNGWithDPI(chart, "./PLOS_AHAH/Figures/Breast_Cancer_Circuit.png", 300);
new SwingWrapper(chart).displayChart();
}