Merge pull request #103 from koosvary/feature/remove-solution-fit-sca…

…ling-in-cli

Change graph export format from PDF to PNG and remove axis scaling

api/src/main/java/org/iconic/ea/strategies/gsemo/GSEMO.java

@@ -70,11 +70,11 @@ public List<R> evolve(List<R> population) {
         // Exactly one parent is required
         final R parent = getSelector(0).apply(newPopulation);
 
-        // Create a single offspring by performing mutation on the parent
+        // Create a single offspring by performing crossover and mutation on the parent
         R offspring = mutate((R) parent.clone());
 
-        // Continue mutating the offspring based on its parent's length
-        // reducing the probability of mutation with each attempt
+        // Continue evolving the offspring based on its parent's length
+        // reducing the probability of evolution with each attempt
         for (int j = 2; j <= parent.getSize(); ++j) {
             if ((1. / (double) j) >= ThreadLocalRandom.current().nextDouble()) {
                 offspring = mutate(offspring);

cli/src/main/java/org/iconic/Client.java

@@ -19,10 +19,10 @@
 import lombok.extern.log4j.Log4j2;
 import org.apache.commons.csv.CSVFormat;
 import org.apache.commons.csv.CSVPrinter;
+import org.iconic.ea.chromosome.cartesian.CartesianChromosome;
 import org.iconic.ea.strategies.EvolutionaryAlgorithm;
 import org.iconic.ea.chromosome.Chromosome;
 import org.iconic.ea.chromosome.ChromosomeFactory;
-import org.iconic.ea.chromosome.cartesian.CartesianChromosome;
 import org.iconic.ea.chromosome.cartesian.CartesianChromosomeFactory;
 import org.iconic.ea.data.DataManager;
 import org.iconic.ea.data.FeatureClass;
@@ -51,6 +51,7 @@
 import java.time.Instant;
 import java.util.*;
 import java.util.List;
+import java.util.concurrent.ThreadLocalRandom;
 import java.util.regex.Pattern;
 import java.util.stream.Collectors;
 
@@ -96,7 +97,7 @@ public static void main(String[] args) {
             log.info("Feature Size: {}", () -> featureSize - 1);
             log.info("Sample Size: {}", () -> sampleSize);
 
-            // Create a supplier for Gene Expression Programming chromosomes
+            // Create a supplier for Gene Cartesian Programming chromosomes
             List<String> inputs = new ArrayList<>(featureSize - 1);
             for (int i = 0; i < featureSize - 1; ++i) {
                 inputs.add(String.valueOf(i));
@@ -116,6 +117,19 @@ public static void main(String[] args) {
                     new Division(), new Power(), new Exponential(), new Sin(), new Cos()
             ));
 
+            final int numConstants = 100;
+            final int min = -100;
+            final int max = 100;
+            final List<FunctionalPrimitive<Double, Double>> constants = new ArrayList<>(numConstants);
+
+            while (constants.size() < numConstants) {
+                constants.add(new Constant<>(ThreadLocalRandom.current().nextDouble(
+                        min, max
+                )));
+            }
+
+            supplier.addFunction(constants);
+
             final int generations = client.getArgs().getGenerations();
             final Set<Chromosome<Double>> nonDominatedFinal = new LinkedHashSet<>();
             final List<Set<Chromosome<Double>>> nonDominatedAll = new ArrayList<>(generations);
@@ -181,6 +195,7 @@ public static void main(String[] args) {
                         graphWriter.write(series.draw());
                     }
 
+                    graphWriter.setAxesLogarithmic(true);
                     graphWriter.export("All Generations - Non-Dominated", directory, "results-all");
                     graphWriter.clear();
 
@@ -192,6 +207,7 @@ public static void main(String[] args) {
 
                     nonDominatedFinal.forEach(seriesWriter::write);
                     graphWriter.write(seriesWriter.draw());
+                    graphWriter.setAxesLogarithmic(true);
                     graphWriter.export("Last Generation - Non-Dominated", directory, "results-last");
 
                     // Create a map of global best values so we can graph just their solution-fit plots
@@ -327,6 +343,9 @@ private static void exportCsv(
                 new FileWriter(new File(directory + "//" + fileName + ".csv")),
                 CSVFormat.EXCEL
         )) {
+            printer.printRecord(
+                    "Mean Squared Error", "Size", "Model"
+            );
             for (final Chromosome<?> chromosome : population) {
                 printer.printRecord(
                         chromosome.getFitness(), chromosome.getSize(), chromosome.toString()
@@ -349,7 +368,7 @@ private static EvolutionaryAlgorithm<CartesianChromosome<Double>, Double> getEvo
         // Create the evolutionary algorithm
         EvolutionaryAlgorithm<CartesianChromosome<Double>, Double> ea =
                 // TODO: generify with a factory so it isn't always GSEMO
-                new GSEMO<>(supplier, 1);
+                new GSEMO<>(supplier, 4);
         ea.setCrossoverProbability(args.getCrossoverProbability());
         ea.setMutationProbability(args.getMutationProbability());
 

cli/src/main/java/org/iconic/utils/GraphWriter.java

@@ -16,6 +16,7 @@
 package org.iconic.utils;
 
 import org.iconic.ea.chromosome.Chromosome;
+import org.knowm.xchart.BitmapEncoder;
 import org.knowm.xchart.VectorGraphicsEncoder;
 import org.knowm.xchart.internal.chartpart.Chart;
 import org.knowm.xchart.internal.series.Series;
@@ -25,6 +26,7 @@
 
 public abstract class GraphWriter<T extends Series> {
     private boolean axesTruncated;
+    private boolean axesLogarithmic;
 
     /**
      * Constructs a new GraphWriter that will truncate outliers from its axes by default.
@@ -48,9 +50,9 @@ public abstract class GraphWriter<T extends Series> {
     public void export(final String chartTitle, final String directory, final String fileName) throws IOException {
         Chart<?, ?> chart = draw();
         chart.setTitle(chartTitle);
-        VectorGraphicsEncoder.saveVectorGraphic(
+        BitmapEncoder.saveBitmap(
                 chart, directory + "//" + fileName,
-                VectorGraphicsEncoder.VectorGraphicsFormat.PDF
+                BitmapEncoder.BitmapFormat.PNG
         );
     }
 
@@ -71,10 +73,24 @@ public boolean isAxesTruncated() {
         return axesTruncated;
     }
 
+    /**
+     * @return True if axes of  the graph written by the GraphWriter will use logarithmic scaling.
+     */
+    public boolean isAxesLogarithmic() {
+        return axesLogarithmic;
+    }
+
     /**
      * @param truncate True if the axes of the graph being written should remove outliers.
      */
     public void setAxesTruncated(boolean truncate) {
         this.axesTruncated = truncate;
     }
+
+    /**
+     * @param logarithmic True if the axes of the graph being written should use logarithmic scaling.
+     */
+    public void setAxesLogarithmic(boolean logarithmic) {
+        this.axesLogarithmic = logarithmic;
+    }
 }

cli/src/main/java/org/iconic/utils/XYGraphWriter.java

@@ -179,7 +179,10 @@ private XYChart getChart() {
                     .theme(Styler.ChartTheme.Matlab).build();
             chart.getStyler().setChartTitleVisible(true);
             chart.getStyler().setMarkerSize(12);
-            chart.getStyler().setXAxisLogarithmic(true);
+
+            if (isAxesLogarithmic()) {
+                chart.getStyler().setXAxisLogarithmic(true);
+            }
         }
 
         return chart;