cleanup of bounding box (#292)

Migrate bounding box code to use float instead of double
aws · Jan 28, 2022 · c76e6aa · c76e6aa
1 parent 6b9d8ad
commit c76e6aa
Show file tree

Hide file tree

Showing 68 changed files with 761 additions and 3,073 deletions.
diff --git a/Java/benchmark/src/main/java/com/amazon/randomcutforest/RandomCutForestBenchmark.java b/Java/benchmark/src/main/java/com/amazon/randomcutforest/RandomCutForestBenchmark.java
@@ -15,8 +15,11 @@
 
 package com.amazon.randomcutforest;
 
-import java.util.Random;
-
+import com.amazon.randomcutforest.returntypes.DensityOutput;
+import com.amazon.randomcutforest.returntypes.DiVector;
+import com.amazon.randomcutforest.returntypes.Neighbor;
+import com.amazon.randomcutforest.testutils.ShingledMultiDimDataWithKeys;
+import org.github.jamm.MemoryMeter;
 import org.openjdk.jmh.annotations.Benchmark;
 import org.openjdk.jmh.annotations.Fork;
 import org.openjdk.jmh.annotations.Level;
@@ -29,9 +32,8 @@
 import org.openjdk.jmh.annotations.Warmup;
 import org.openjdk.jmh.infra.Blackhole;
 
-import com.amazon.randomcutforest.returntypes.DensityOutput;
-import com.amazon.randomcutforest.returntypes.DiVector;
-import com.amazon.randomcutforest.testutils.NormalMixtureTestData;
+import java.util.List;
+import java.util.Random;
 
 @Warmup(iterations = 2)
 @Measurement(iterations = 5)
@@ -43,28 +45,38 @@ public class RandomCutForestBenchmark {
 
     @State(Scope.Benchmark)
     public static class BenchmarkState {
-        @Param({ "10" })
-        int dimensions;
+        @Param({ "5" })
+        int baseDimensions;
 
-        @Param({ "50" })
+        @Param({ "8" })
+        int shingleSize;
+
+        @Param({ "30" })
         int numberOfTrees;
 
+        @Param({ "1.0", "0.9", "0.8", "0.7", "0.6", "0.5", "0.4", "0.3", "0.2", "0.1", "0.0" })
+        double boundingBoxCacheFraction;
+
         @Param({ "false", "true" })
-        boolean compact;
+        boolean parallel;
 
         double[][] data;
         RandomCutForest forest;
 
         @Setup(Level.Trial)
         public void setUpData() {
-            NormalMixtureTestData testData = new NormalMixtureTestData();
-            data = testData.generateTestData(DATA_SIZE, dimensions);
+            int dimensions = baseDimensions * shingleSize;
+            int sampleSize = 256;
+            int dataSize = 100 * sampleSize;
+            data = ShingledMultiDimDataWithKeys.getMultiDimData(dataSize + shingleSize - 1, 50, 100, 5, 17,
+                    baseDimensions).data;
         }
 
         @Setup(Level.Invocation)
         public void setUpForest() {
-            forest = RandomCutForest.builder().numberOfTrees(numberOfTrees).dimensions(dimensions)
-                    .parallelExecutionEnabled(false).compact(compact).randomSeed(99).build();
+            forest = RandomCutForest.builder().numberOfTrees(numberOfTrees).dimensions(baseDimensions * shingleSize)
+                    .internalShinglingEnabled(true).shingleSize(shingleSize).parallelExecutionEnabled(parallel)
+                    .boundingBoxCacheFraction(boundingBoxCacheFraction).randomSeed(99).build();
         }
     }
 
@@ -115,6 +127,10 @@ public RandomCutForest scoreAndUpdate(BenchmarkState state, Blackhole blackhole)
         }
 
         blackhole.consume(score);
+        if (!forest.parallelExecutionEnabled) {
+            MemoryMeter meter = new MemoryMeter();
+            System.out.println(" forest size " + meter.measureDeep(forest));
+        }
         return forest;
     }
 
@@ -149,4 +165,36 @@ public RandomCutForest basicDensityAndUpdate(BenchmarkState state, Blackhole bla
         blackhole.consume(output);
         return forest;
     }
+
+    @Benchmark
+    @OperationsPerInvocation(DATA_SIZE)
+    public RandomCutForest basicNeighborAndUpdate(BenchmarkState state, Blackhole blackhole) {
+        double[][] data = state.data;
+        forest = state.forest;
+        List<Neighbor> output = null;
+
+        for (int i = 0; i < data.length; i++) {
+            output = forest.getNearNeighborsInSample(data[i]);
+            forest.update(data[i]);
+        }
+
+        blackhole.consume(output);
+        return forest;
+    }
+
+    @Benchmark
+    @OperationsPerInvocation(DATA_SIZE)
+    public RandomCutForest basicExtrapolateAndUpdate(BenchmarkState state, Blackhole blackhole) {
+        double[][] data = state.data;
+        forest = state.forest;
+        double[] output = null;
+
+        for (int i = 0; i < data.length; i++) {
+            output = forest.extrapolate(1);
+            forest.update(data[i]);
+        }
+
+        blackhole.consume(output);
+        return forest;
+    }
 }
diff --git a/Java/core/src/main/java/com/amazon/randomcutforest/CommonUtils.java b/Java/core/src/main/java/com/amazon/randomcutforest/CommonUtils.java
@@ -100,7 +100,7 @@ public static <T> T checkNotNull(T object, String message) {
      * @return the probability of separation choosing a random cut
      */
 
-    public static double getProbabilityOfSeparation(final IBoundingBoxView boundingBox, double[] queryPoint) {
+    public static double getProbabilityOfSeparation(final IBoundingBoxView boundingBox, float[] queryPoint) {
         double sumOfNewRange = 0d;
         double sumOfDifferenceInRange = 0d;
 

diff --git a/Java/core/src/main/java/com/amazon/randomcutforest/IMultiVisitorFactory.java b/Java/core/src/main/java/com/amazon/randomcutforest/IMultiVisitorFactory.java
@@ -19,7 +19,7 @@
 
 @FunctionalInterface
 public interface IMultiVisitorFactory<R> {
-    MultiVisitor<R> newVisitor(ITree<?, ?> tree, double[] point);
+    MultiVisitor<R> newVisitor(ITree<?, ?> tree, float[] point);
 
     default R liftResult(ITree<?, ?> tree, R result) {
         return result;

diff --git a/Java/core/src/main/java/com/amazon/randomcutforest/IVisitorFactory.java b/Java/core/src/main/java/com/amazon/randomcutforest/IVisitorFactory.java
@@ -19,7 +19,7 @@
 
 @FunctionalInterface
 public interface IVisitorFactory<R> {
-    Visitor<R> newVisitor(ITree<?, ?> tree, double[] point);
+    Visitor<R> newVisitor(ITree<?, ?> tree, float[] point);
 
     default R liftResult(ITree<?, ?> tree, R result) {
         return result;

diff --git a/Java/core/src/main/java/com/amazon/randomcutforest/MultiVisitorFactory.java b/Java/core/src/main/java/com/amazon/randomcutforest/MultiVisitorFactory.java
@@ -29,21 +29,21 @@
  */
 
 public class MultiVisitorFactory<R> implements IMultiVisitorFactory<R> {
-    private final BiFunction<ITree<?, ?>, double[], MultiVisitor<R>> newVisitor;
+    private final BiFunction<ITree<?, ?>, float[], MultiVisitor<R>> newVisitor;
     private final BiFunction<ITree<?, ?>, R, R> liftResult;
 
-    public MultiVisitorFactory(BiFunction<ITree<?, ?>, double[], MultiVisitor<R>> newVisitor,
+    public MultiVisitorFactory(BiFunction<ITree<?, ?>, float[], MultiVisitor<R>> newVisitor,
             BiFunction<ITree<?, ?>, R, R> liftResult) {
         this.newVisitor = newVisitor;
         this.liftResult = liftResult;
     }
 
-    public MultiVisitorFactory(BiFunction<ITree<?, ?>, double[], MultiVisitor<R>> newVisitor) {
+    public MultiVisitorFactory(BiFunction<ITree<?, ?>, float[], MultiVisitor<R>> newVisitor) {
         this(newVisitor, (tree, x) -> x);
     }
 
     @Override
-    public MultiVisitor<R> newVisitor(ITree<?, ?> tree, double[] point) {
+    public MultiVisitor<R> newVisitor(ITree<?, ?> tree, float[] point) {
         return newVisitor.apply(tree, point);
     }