Skip to content

Commit

Permalink
IGNITE-7829: Adopt kNN regression example to the new Partitioned Dataset
Browse files Browse the repository at this point in the history 
this closes #3798
  • Loading branch information
@zaleslaw @YuriBabak
zaleslaw authored and YuriBabak committed Apr 13, 2018
1 parent 9be3357 commit 8550d61
Show file tree
Hide file tree
Showing 18 changed files with 453 additions and 243 deletions.
4 changes: 2 additions & 2 deletions examples/src/main/java/org/apache/ignite/examples/ml/knn/KNNClassificationExample.java
View file
Expand Up @@ -80,7 +80,7 @@ public static void main(String[] args) throws InterruptedException {
double prediction = knnMdl.apply(new DenseLocalOnHeapVector(inputs));

totalAmount++;
if(groundTruth != prediction)
if (groundTruth != prediction)
amountOfErrors++;

System.out.printf(">>> | %.4f\t\t| %.4f\t\t|\n", prediction, groundTruth);
Expand All @@ -89,7 +89,7 @@ public static void main(String[] args) throws InterruptedException {
System.out.println(">>> ---------------------------------");

System.out.println("\n>>> Absolute amount of errors " + amountOfErrors);
System.out.println("\n>>> Accuracy " + (1 - amountOfErrors / (double)totalAmount));
System.out.println("\n>>> Accuracy " + (1 - amountOfErrors / (double) totalAmount));
}
});

Expand Down
310 changes: 310 additions & 0 deletions examples/src/main/java/org/apache/ignite/examples/ml/knn/KNNRegressionExample.java
View file
@@ -0,0 +1,310 @@
/*
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership.
* The ASF licenses this file to You under the Apache License, Version 2.0
* (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/

package org.apache.ignite.examples.ml.knn;

import java.util.Arrays;
import java.util.UUID;
import javax.cache.Cache;
import org.apache.ignite.Ignite;
import org.apache.ignite.IgniteCache;
import org.apache.ignite.Ignition;
import org.apache.ignite.cache.affinity.rendezvous.RendezvousAffinityFunction;
import org.apache.ignite.cache.query.QueryCursor;
import org.apache.ignite.cache.query.ScanQuery;
import org.apache.ignite.configuration.CacheConfiguration;
import org.apache.ignite.ml.dataset.impl.cache.CacheBasedDatasetBuilder;
import org.apache.ignite.ml.knn.classification.KNNClassificationTrainer;
import org.apache.ignite.ml.knn.classification.KNNStrategy;
import org.apache.ignite.ml.knn.regression.KNNRegressionModel;
import org.apache.ignite.ml.knn.regression.KNNRegressionTrainer;
import org.apache.ignite.ml.math.distances.ManhattanDistance;
import org.apache.ignite.ml.math.impls.vector.DenseLocalOnHeapVector;
import org.apache.ignite.thread.IgniteThread;

/**
* Run kNN regression trainer over distributed dataset.
*
* @see KNNClassificationTrainer
*/
public class KNNRegressionExample {
/** Run example. */
public static void main(String[] args) throws InterruptedException {
System.out.println();
System.out.println(">>> kNN regression over cached dataset usage example started.");
// Start ignite grid.
try (Ignite ignite = Ignition.start("examples/config/example-ignite.xml")) {
System.out.println(">>> Ignite grid started.");

IgniteThread igniteThread = new IgniteThread(ignite.configuration().getIgniteInstanceName(),
KNNRegressionExample.class.getSimpleName(), () -> {
IgniteCache<Integer, double[]> dataCache = getTestCache(ignite);

KNNRegressionTrainer trainer = new KNNRegressionTrainer();

KNNRegressionModel knnMdl = (KNNRegressionModel) trainer.fit(
new CacheBasedDatasetBuilder<>(ignite, dataCache),
(k, v) -> Arrays.copyOfRange(v, 1, v.length),
(k, v) -> v[0]
).withK(5)
.withDistanceMeasure(new ManhattanDistance())
.withStrategy(KNNStrategy.WEIGHTED);

int totalAmount = 0;
// Calculate mean squared error (MSE)
double mse = 0.0;
// Calculate mean absolute error (MAE)
double mae = 0.0;

try (QueryCursor<Cache.Entry<Integer, double[]>> observations = dataCache.query(new ScanQuery<>())) {
for (Cache.Entry<Integer, double[]> observation : observations) {
double[] val = observation.getValue();
double[] inputs = Arrays.copyOfRange(val, 1, val.length);
double groundTruth = val[0];

double prediction = knnMdl.apply(new DenseLocalOnHeapVector(inputs));

mse += Math.pow(prediction - groundTruth, 2.0);
mae += Math.abs(prediction - groundTruth);

totalAmount++;
}

mse = mse / totalAmount;
System.out.println("\n>>> Mean squared error (MSE) " + mse);

mae = mae / totalAmount;
System.out.println("\n>>> Mean absolute error (MAE) " + mae);
}
});

igniteThread.start();
igniteThread.join();
}
}

/**
* Fills cache with data and returns it.
*
* @param ignite Ignite instance.
* @return Filled Ignite Cache.
*/
private static IgniteCache<Integer, double[]> getTestCache(Ignite ignite) {
CacheConfiguration<Integer, double[]> cacheConfiguration = new CacheConfiguration<>();
cacheConfiguration.setName("TEST_" + UUID.randomUUID());
cacheConfiguration.setAffinity(new RendezvousAffinityFunction(false, 10));

IgniteCache<Integer, double[]> cache = ignite.createCache(cacheConfiguration);

for (int i = 0; i < data.length; i++)
cache.put(i, data[i]);

return cache;
}

/** The Iris dataset. */
private static final double[][] data = {
{199, 125, 256, 6000, 256, 16, 128},
{253, 29, 8000, 32000, 32, 8, 32},
{132, 29, 8000, 16000, 32, 8, 16},
{290, 26, 8000, 32000, 64, 8, 32},
{381, 23, 16000, 32000, 64, 16, 32},
{749, 23, 16000, 64000, 64, 16, 32},
{1238, 23, 32000, 64000, 128, 32, 64},
{23, 400, 1000, 3000, 0, 1, 2},
{24, 400, 512, 3500, 4, 1, 6},
{70, 60, 2000, 8000, 65, 1, 8},
{117, 50, 4000, 16000, 65, 1, 8},
{15, 350, 64, 64, 0, 1, 4},
{64, 200, 512, 16000, 0, 4, 32},
{23, 167, 524, 2000, 8, 4, 15},
{29, 143, 512, 5000, 0, 7, 32},
{22, 143, 1000, 2000, 0, 5, 16},
{124, 110, 5000, 5000, 142, 8, 64},
{35, 143, 1500, 6300, 0, 5, 32},
{39, 143, 3100, 6200, 0, 5, 20},
{40, 143, 2300, 6200, 0, 6, 64},
{45, 110, 3100, 6200, 0, 6, 64},
{28, 320, 128, 6000, 0, 1, 12},
{21, 320, 512, 2000, 4, 1, 3},
{28, 320, 256, 6000, 0, 1, 6},
{22, 320, 256, 3000, 4, 1, 3},
{28, 320, 512, 5000, 4, 1, 5},
{27, 320, 256, 5000, 4, 1, 6},
{102, 25, 1310, 2620, 131, 12, 24},
{74, 50, 2620, 10480, 30, 12, 24},
{138, 56, 5240, 20970, 30, 12, 24},
{136, 64, 5240, 20970, 30, 12, 24},
{23, 50, 500, 2000, 8, 1, 4},
{29, 50, 1000, 4000, 8, 1, 5},
{44, 50, 2000, 8000, 8, 1, 5},
{30, 50, 1000, 4000, 8, 3, 5},
{41, 50, 1000, 8000, 8, 3, 5},
{74, 50, 2000, 16000, 8, 3, 5},
{54, 133, 1000, 12000, 9, 3, 12},
{41, 133, 1000, 8000, 9, 3, 12},
{18, 810, 512, 512, 8, 1, 1},
{28, 810, 1000, 5000, 0, 1, 1},
{36, 320, 512, 8000, 4, 1, 5},
{38, 200, 512, 8000, 8, 1, 8},
{34, 700, 384, 8000, 0, 1, 1},
{19, 700, 256, 2000, 0, 1, 1},
{72, 140, 1000, 16000, 16, 1, 3},
{36, 200, 1000, 8000, 0, 1, 2},
{30, 110, 1000, 4000, 16, 1, 2},
{56, 110, 1000, 12000, 16, 1, 2},
{42, 220, 1000, 8000, 16, 1, 2},
{34, 800, 256, 8000, 0, 1, 4},
{19, 125, 512, 1000, 0, 8, 20},
{75, 75, 2000, 8000, 64, 1, 38},
{113, 75, 2000, 16000, 64, 1, 38},
{157, 75, 2000, 16000, 128, 1, 38},
{18, 90, 256, 1000, 0, 3, 10},
{20, 105, 256, 2000, 0, 3, 10},
{28, 105, 1000, 4000, 0, 3, 24},
{33, 105, 2000, 4000, 8, 3, 19},
{47, 75, 2000, 8000, 8, 3, 24},
{54, 75, 3000, 8000, 8, 3, 48},
{20, 175, 256, 2000, 0, 3, 24},
{23, 300, 768, 3000, 0, 6, 24},
{25, 300, 768, 3000, 6, 6, 24},
{52, 300, 768, 12000, 6, 6, 24},
{27, 300, 768, 4500, 0, 1, 24},
{50, 300, 384, 12000, 6, 1, 24},
{18, 300, 192, 768, 6, 6, 24},
{53, 180, 768, 12000, 6, 1, 31},
{23, 330, 1000, 3000, 0, 2, 4},
{30, 300, 1000, 4000, 8, 3, 64},
{73, 300, 1000, 16000, 8, 2, 112},
{20, 330, 1000, 2000, 0, 1, 2},
{25, 330, 1000, 4000, 0, 3, 6},
{28, 140, 2000, 4000, 0, 3, 6},
{29, 140, 2000, 4000, 0, 4, 8},
{32, 140, 2000, 4000, 8, 1, 20},
{175, 140, 2000, 32000, 32, 1, 20},
{57, 140, 2000, 8000, 32, 1, 54},
{181, 140, 2000, 32000, 32, 1, 54},
{32, 140, 2000, 4000, 8, 1, 20},
{82, 57, 4000, 16000, 1, 6, 12},
{171, 57, 4000, 24000, 64, 12, 16},
{361, 26, 16000, 32000, 64, 16, 24},
{350, 26, 16000, 32000, 64, 8, 24},
{220, 26, 8000, 32000, 0, 8, 24},
{113, 26, 8000, 16000, 0, 8, 16},
{15, 480, 96, 512, 0, 1, 1},
{21, 203, 1000, 2000, 0, 1, 5},
{35, 115, 512, 6000, 16, 1, 6},
{18, 1100, 512, 1500, 0, 1, 1},
{20, 1100, 768, 2000, 0, 1, 1},
{20, 600, 768, 2000, 0, 1, 1},
{28, 400, 2000, 4000, 0, 1, 1},
{45, 400, 4000, 8000, 0, 1, 1},
{18, 900, 1000, 1000, 0, 1, 2},
{17, 900, 512, 1000, 0, 1, 2},
{26, 900, 1000, 4000, 4, 1, 2},
{28, 900, 1000, 4000, 8, 1, 2},
{28, 900, 2000, 4000, 0, 3, 6},
{31, 225, 2000, 4000, 8, 3, 6},
{42, 180, 2000, 8000, 8, 1, 6},
{76, 185, 2000, 16000, 16, 1, 6},
{76, 180, 2000, 16000, 16, 1, 6},
{26, 225, 1000, 4000, 2, 3, 6},
{59, 25, 2000, 12000, 8, 1, 4},
{65, 25, 2000, 12000, 16, 3, 5},
{101, 17, 4000, 16000, 8, 6, 12},
{116, 17, 4000, 16000, 32, 6, 12},
{18, 1500, 768, 1000, 0, 0, 0},
{20, 1500, 768, 2000, 0, 0, 0},
{20, 800, 768, 2000, 0, 0, 0},
{30, 50, 2000, 4000, 0, 3, 6},
{44, 50, 2000, 8000, 8, 3, 6},
{82, 50, 2000, 16000, 24, 1, 6},
{128, 50, 8000, 16000, 48, 1, 10},
{37, 100, 1000, 8000, 0, 2, 6},
{46, 100, 1000, 8000, 24, 2, 6},
{46, 100, 1000, 8000, 24, 3, 6},
{80, 50, 2000, 16000, 12, 3, 16},
{88, 50, 2000, 16000, 24, 6, 16},
{33, 150, 512, 4000, 0, 8, 128},
{46, 115, 2000, 8000, 16, 1, 3},
{29, 115, 2000, 4000, 2, 1, 5},
{53, 92, 2000, 8000, 32, 1, 6},
{41, 92, 2000, 8000, 4, 1, 6},
{86, 75, 4000, 16000, 16, 1, 6},
{95, 60, 4000, 16000, 32, 1, 6},
{107, 60, 2000, 16000, 64, 5, 8},
{117, 60, 4000, 16000, 64, 5, 8},
{119, 50, 4000, 16000, 64, 5, 10},
{120, 72, 4000, 16000, 64, 8, 16},
{48, 72, 2000, 8000, 16, 6, 8},
{126, 40, 8000, 16000, 32, 8, 16},
{266, 40, 8000, 32000, 64, 8, 24},
{270, 35, 8000, 32000, 64, 8, 24},
{426, 38, 16000, 32000, 128, 16, 32},
{151, 48, 4000, 24000, 32, 8, 24},
{267, 38, 8000, 32000, 64, 8, 24},
{603, 30, 16000, 32000, 256, 16, 24},
{19, 112, 1000, 1000, 0, 1, 4},
{21, 84, 1000, 2000, 0, 1, 6},
{26, 56, 1000, 4000, 0, 1, 6},
{35, 56, 2000, 6000, 0, 1, 8},
{41, 56, 2000, 8000, 0, 1, 8},
{47, 56, 4000, 8000, 0, 1, 8},
{62, 56, 4000, 12000, 0, 1, 8},
{78, 56, 4000, 16000, 0, 1, 8},
{80, 38, 4000, 8000, 32, 16, 32},
{142, 38, 8000, 16000, 64, 4, 8},
{281, 38, 8000, 24000, 160, 4, 8},
{190, 38, 4000, 16000, 128, 16, 32},
{21, 200, 1000, 2000, 0, 1, 2},
{25, 200, 1000, 4000, 0, 1, 4},
{67, 200, 2000, 8000, 64, 1, 5},
{24, 250, 512, 4000, 0, 1, 7},
{24, 250, 512, 4000, 0, 4, 7},
{64, 250, 1000, 16000, 1, 1, 8},
{25, 160, 512, 4000, 2, 1, 5},
{20, 160, 512, 2000, 2, 3, 8},
{29, 160, 1000, 4000, 8, 1, 14},
{43, 160, 1000, 8000, 16, 1, 14},
{53, 160, 2000, 8000, 32, 1, 13},
{19, 240, 512, 1000, 8, 1, 3},
{22, 240, 512, 2000, 8, 1, 5},
{31, 105, 2000, 4000, 8, 3, 8},
{41, 105, 2000, 6000, 16, 6, 16},
{47, 105, 2000, 8000, 16, 4, 14},
{99, 52, 4000, 16000, 32, 4, 12},
{67, 70, 4000, 12000, 8, 6, 8},
{81, 59, 4000, 12000, 32, 6, 12},
{149, 59, 8000, 16000, 64, 12, 24},
{183, 26, 8000, 24000, 32, 8, 16},
{275, 26, 8000, 32000, 64, 12, 16},
{382, 26, 8000, 32000, 128, 24, 32},
{56, 116, 2000, 8000, 32, 5, 28},
{182, 50, 2000, 32000, 24, 6, 26},
{227, 50, 2000, 32000, 48, 26, 52},
{341, 50, 2000, 32000, 112, 52, 104},
{360, 50, 4000, 32000, 112, 52, 104},
{919, 30, 8000, 64000, 96, 12, 176},
{978, 30, 8000, 64000, 128, 12, 176},
{24, 180, 262, 4000, 0, 1, 3},
{37, 124, 1000, 8000, 0, 1, 8},
{50, 98, 1000, 8000, 32, 2, 8},
{41, 125, 2000, 8000, 0, 2, 14},
{47, 480, 512, 8000, 32, 0, 0},
{25, 480, 1000, 4000, 0, 0, 0}
};
}
10 changes: 5 additions & 5 deletions modules/ml/src/main/java/org/apache/ignite/ml/knn/KNNUtils.java
View file
Expand Up @@ -20,7 +20,7 @@
import org.apache.ignite.ml.dataset.Dataset;
import org.apache.ignite.ml.dataset.DatasetBuilder;
import org.apache.ignite.ml.dataset.PartitionDataBuilder;
import org.apache.ignite.ml.knn.partitions.KNNPartitionContext;
import org.apache.ignite.ml.dataset.primitive.context.EmptyContext;
import org.apache.ignite.ml.math.functions.IgniteBiFunction;
import org.apache.ignite.ml.structures.LabeledDataset;
import org.apache.ignite.ml.structures.LabeledVector;
Expand All @@ -39,18 +39,18 @@ public class KNNUtils {
* @param lbExtractor Label extractor.
* @return Dataset.
*/
@Nullable public static <K, V> Dataset<KNNPartitionContext, LabeledDataset<Double, LabeledVector>> buildDataset(DatasetBuilder<K, V> datasetBuilder, IgniteBiFunction<K, V, double[]> featureExtractor, IgniteBiFunction<K, V, Double> lbExtractor) {
PartitionDataBuilder<K, V, KNNPartitionContext, LabeledDataset<Double, LabeledVector>> partDataBuilder
@Nullable public static <K, V> Dataset<EmptyContext, LabeledDataset<Double, LabeledVector>> buildDataset(DatasetBuilder<K, V> datasetBuilder, IgniteBiFunction<K, V, double[]> featureExtractor, IgniteBiFunction<K, V, Double> lbExtractor) {
PartitionDataBuilder<K, V, EmptyContext, LabeledDataset<Double, LabeledVector>> partDataBuilder
= new LabeledDatasetPartitionDataBuilderOnHeap<>(
featureExtractor,
lbExtractor
);

Dataset<KNNPartitionContext, LabeledDataset<Double, LabeledVector>> dataset = null;
Dataset<EmptyContext, LabeledDataset<Double, LabeledVector>> dataset = null;

if (datasetBuilder != null) {
dataset = datasetBuilder.build(
(upstream, upstreamSize) -> new KNNPartitionContext(),
(upstream, upstreamSize) -> new EmptyContext(),
partDataBuilder
);
}
Expand Down
9 changes: 6 additions & 3 deletions modules/ml/src/main/java/org/apache/ignite/ml/knn/classification/KNNClassificationModel.java
View file
Expand Up @@ -32,7 +32,7 @@
import org.apache.ignite.ml.Exporter;
import org.apache.ignite.ml.Model;
import org.apache.ignite.ml.dataset.Dataset;
import org.apache.ignite.ml.knn.partitions.KNNPartitionContext;
import org.apache.ignite.ml.dataset.primitive.context.EmptyContext;
import org.apache.ignite.ml.math.Vector;
import org.apache.ignite.ml.math.distances.DistanceMeasure;
import org.apache.ignite.ml.math.distances.EuclideanDistance;
Expand All @@ -44,6 +44,9 @@
* kNN algorithm model to solve multi-class classification task.
*/
public class KNNClassificationModel<K, V> implements Model<Vector, Double>, Exportable<KNNModelFormat> {
/** */
private static final long serialVersionUID = -127386523291350345L;

/** Amount of nearest neighbors. */
protected int k = 5;

Expand All @@ -54,13 +57,13 @@ public class KNNClassificationModel<K, V> implements Model<Vector, Double>, Expo
protected KNNStrategy stgy = KNNStrategy.SIMPLE;

/** Dataset. */
private Dataset<KNNPartitionContext, LabeledDataset<Double, LabeledVector>> dataset;
private Dataset<EmptyContext, LabeledDataset<Double, LabeledVector>> dataset;

/**
* Builds the model via prepared dataset.
* @param dataset Specially prepared object to run algorithm over it.
*/
public KNNClassificationModel(Dataset<KNNPartitionContext, LabeledDataset<Double, LabeledVector>> dataset) {
public KNNClassificationModel(Dataset<EmptyContext, LabeledDataset<Double, LabeledVector>> dataset) {
this.dataset = dataset;
}

Expand Down

0 comments on commit 8550d61

Please sign in to comment.