yunfengzhou-hub commented on code in PR #148: URL: https://github.com/apache/flink-ml/pull/148#discussion_r960171818
########## flink-ml-lib/src/test/java/org/apache/flink/ml/clustering/AgglomerativeClusteringTest.java: ########## @@ -0,0 +1,312 @@ +/* + * 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.flink.ml.clustering; + +import org.apache.flink.api.common.restartstrategy.RestartStrategies; +import org.apache.flink.configuration.Configuration; +import org.apache.flink.ml.clustering.agglomerativeclustering.AgglomerativeClustering; +import org.apache.flink.ml.clustering.agglomerativeclustering.AgglomerativeClusteringParams; +import org.apache.flink.ml.common.distance.CosineDistanceMeasure; +import org.apache.flink.ml.common.distance.EuclideanDistanceMeasure; +import org.apache.flink.ml.common.distance.ManhattanDistanceMeasure; +import org.apache.flink.ml.linalg.DenseVector; +import org.apache.flink.ml.linalg.Vectors; +import org.apache.flink.ml.util.TestUtils; +import org.apache.flink.streaming.api.environment.ExecutionCheckpointingOptions; +import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment; +import org.apache.flink.table.api.Table; +import org.apache.flink.table.api.bridge.java.StreamTableEnvironment; +import org.apache.flink.test.util.AbstractTestBase; +import org.apache.flink.types.Row; + +import org.apache.commons.collections.IteratorUtils; +import org.junit.Before; +import org.junit.Rule; +import org.junit.Test; +import org.junit.rules.TemporaryFolder; + +import java.util.Arrays; +import java.util.Comparator; +import java.util.List; + +import static org.junit.Assert.assertEquals; +import static org.junit.Assert.assertFalse; +import static org.junit.Assert.assertNull; +import static org.junit.Assert.assertTrue; + +/** Tests {@link AgglomerativeClustering}. */ +public class AgglomerativeClusteringTest extends AbstractTestBase { + + @Rule public final TemporaryFolder tempFolder = new TemporaryFolder(); + private StreamTableEnvironment tEnv; + private StreamExecutionEnvironment env; + private Table inputDataTable; + + private static final List<DenseVector> INPUT_DATA = + Arrays.asList( + Vectors.dense(1, 1), + Vectors.dense(1, 4), + Vectors.dense(1, 0), + Vectors.dense(4, 1.5), + Vectors.dense(4, 4), + Vectors.dense(4, 0)); + + private static final List<Row> EUCLIDEAN_AVERAGE_MERGE_INFO = + Arrays.asList( + Row.of(0, 2, 1, 2), + Row.of(3, 5, 1.5, 2), + Row.of(1, 4, 3, 2), + Row.of(6, 7, 3.1394402, 4), + Row.of(8, 9, 3.9559706, 6)); + + private static final List<Row> COSINE_AVERAGE_MERGE_INFO = + Arrays.asList( + Row.of(2, 5, 0, 2), + Row.of(0, 4, 1.1102230E-16, 2), + Row.of(3, 6, 0.0636708, 3), + Row.of(1, 7, 0.1425070, 3), + Row.of(8, 9, 0.3664484, 6)); + + private static final List<Row> MANHATTAN_AVERAGE_MERGE_INFO = + Arrays.asList( + Row.of(0, 2, 1, 2), + Row.of(3, 5, 1.5, 2), + Row.of(1, 4, 3, 2), + Row.of(6, 7, 3.75, 4), + Row.of(8, 9, 4.875, 6)); + + private static final List<Row> EUCLIDEAN_SINGLE_MERGE_INFO = + Arrays.asList( + Row.of(0, 2, 1, 2), + Row.of(3, 5, 1.5, 2), + Row.of(4, 7, 2.5, 3), + Row.of(8, 9, 3, 6), + Row.of(1, 6, 3, 3)); + + private static final List<Row> EUCLIDEAN_WARD_MERGE_INFO = + Arrays.asList( + Row.of(0, 2, 1, 2), + Row.of(3, 5, 1.5, 2), + Row.of(1, 4, 3, 2), + Row.of(6, 7, 4.2573465, 4), + Row.of(8, 9, 5.5113519, 6)); + + private static final List<Row> EUCLIDEAN_COMPLETE_MERGE_INFO = + Arrays.asList( + Row.of(0, 2, 1, 2), + Row.of(3, 5, 1.5, 2), + Row.of(1, 4, 3, 2), + Row.of(6, 7, 3.3541019, 4), + Row.of(8, 9, 5, 6)); + + private static final List<Row> EUCLIDEAN_WARD_K_AS_TWO_RESULT = + Arrays.asList( + Row.of(Vectors.dense(1, 1), 0), + Row.of(Vectors.dense(1, 4), 1), + Row.of(Vectors.dense(1, 0), 0), + Row.of(Vectors.dense(4, 1.5), 0), + Row.of(Vectors.dense(4, 4), 1), + Row.of(Vectors.dense(4, 0), 0)); + + private static final List<Row> EUCLIDEAN_WARD_THRESHOLD_AS_TWO_RESULT = + Arrays.asList( + Row.of(Vectors.dense(1, 1), 0), + Row.of(Vectors.dense(1, 4), 1), + Row.of(Vectors.dense(1, 0), 0), + Row.of(Vectors.dense(4, 1.5), 2), + Row.of(Vectors.dense(4, 4), 1), + Row.of(Vectors.dense(4, 0), 2)); + + private static final double TOLERANCE = 1e-7; + + @Before + public void before() { + Configuration config = new Configuration(); + config.set(ExecutionCheckpointingOptions.ENABLE_CHECKPOINTS_AFTER_TASKS_FINISH, true); + env = StreamExecutionEnvironment.getExecutionEnvironment(config); + env.setParallelism(3); + env.enableCheckpointing(100); + env.setRestartStrategy(RestartStrategies.noRestart()); + tEnv = StreamTableEnvironment.create(env); + inputDataTable = tEnv.fromDataStream(env.fromCollection(INPUT_DATA)).as("features"); Review Comment: If the input data comes from an operator whose parallelism is 3, how can we make sure that they will arrive at the `LocalAgglomerativeClusteringOperator`, whose parallelism is 1, in the same order and be assigned the same cluster ids across difference runs? If each record might be assigned a different id in different runs, why would the merge info remains the same and can be verified with an expected value? ########## flink-ml-lib/src/test/java/org/apache/flink/ml/clustering/AgglomerativeClusteringTest.java: ########## @@ -0,0 +1,312 @@ +/* + * 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.flink.ml.clustering; + +import org.apache.flink.api.common.restartstrategy.RestartStrategies; +import org.apache.flink.configuration.Configuration; +import org.apache.flink.ml.clustering.agglomerativeclustering.AgglomerativeClustering; +import org.apache.flink.ml.clustering.agglomerativeclustering.AgglomerativeClusteringParams; +import org.apache.flink.ml.common.distance.CosineDistanceMeasure; +import org.apache.flink.ml.common.distance.EuclideanDistanceMeasure; +import org.apache.flink.ml.common.distance.ManhattanDistanceMeasure; +import org.apache.flink.ml.linalg.DenseVector; +import org.apache.flink.ml.linalg.Vectors; +import org.apache.flink.ml.util.TestUtils; +import org.apache.flink.streaming.api.environment.ExecutionCheckpointingOptions; +import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment; +import org.apache.flink.table.api.Table; +import org.apache.flink.table.api.bridge.java.StreamTableEnvironment; +import org.apache.flink.test.util.AbstractTestBase; +import org.apache.flink.types.Row; + +import org.apache.commons.collections.IteratorUtils; +import org.junit.Before; +import org.junit.Rule; +import org.junit.Test; +import org.junit.rules.TemporaryFolder; + +import java.util.Arrays; +import java.util.Comparator; +import java.util.List; + +import static org.junit.Assert.assertEquals; +import static org.junit.Assert.assertFalse; +import static org.junit.Assert.assertNull; +import static org.junit.Assert.assertTrue; + +/** Tests {@link AgglomerativeClustering}. */ +public class AgglomerativeClusteringTest extends AbstractTestBase { + + @Rule public final TemporaryFolder tempFolder = new TemporaryFolder(); + private StreamTableEnvironment tEnv; + private StreamExecutionEnvironment env; + private Table inputDataTable; + + private static final List<DenseVector> INPUT_DATA = + Arrays.asList( + Vectors.dense(1, 1), + Vectors.dense(1, 4), + Vectors.dense(1, 0), + Vectors.dense(4, 1.5), + Vectors.dense(4, 4), + Vectors.dense(4, 0)); + + private static final List<Row> EUCLIDEAN_AVERAGE_MERGE_INFO = + Arrays.asList( + Row.of(0, 2, 1, 2), + Row.of(3, 5, 1.5, 2), + Row.of(1, 4, 3, 2), + Row.of(6, 7, 3.1394402, 4), + Row.of(8, 9, 3.9559706, 6)); + + private static final List<Row> COSINE_AVERAGE_MERGE_INFO = + Arrays.asList( + Row.of(2, 5, 0, 2), + Row.of(0, 4, 1.1102230E-16, 2), + Row.of(3, 6, 0.0636708, 3), + Row.of(1, 7, 0.1425070, 3), + Row.of(8, 9, 0.3664484, 6)); + + private static final List<Row> MANHATTAN_AVERAGE_MERGE_INFO = + Arrays.asList( + Row.of(0, 2, 1, 2), + Row.of(3, 5, 1.5, 2), + Row.of(1, 4, 3, 2), + Row.of(6, 7, 3.75, 4), + Row.of(8, 9, 4.875, 6)); + + private static final List<Row> EUCLIDEAN_SINGLE_MERGE_INFO = + Arrays.asList( + Row.of(0, 2, 1, 2), + Row.of(3, 5, 1.5, 2), + Row.of(4, 7, 2.5, 3), + Row.of(8, 9, 3, 6), + Row.of(1, 6, 3, 3)); + + private static final List<Row> EUCLIDEAN_WARD_MERGE_INFO = + Arrays.asList( + Row.of(0, 2, 1, 2), + Row.of(3, 5, 1.5, 2), + Row.of(1, 4, 3, 2), + Row.of(6, 7, 4.2573465, 4), + Row.of(8, 9, 5.5113519, 6)); + + private static final List<Row> EUCLIDEAN_COMPLETE_MERGE_INFO = + Arrays.asList( + Row.of(0, 2, 1, 2), + Row.of(3, 5, 1.5, 2), + Row.of(1, 4, 3, 2), + Row.of(6, 7, 3.3541019, 4), + Row.of(8, 9, 5, 6)); + + private static final List<Row> EUCLIDEAN_WARD_K_AS_TWO_RESULT = + Arrays.asList( + Row.of(Vectors.dense(1, 1), 0), + Row.of(Vectors.dense(1, 4), 1), + Row.of(Vectors.dense(1, 0), 0), + Row.of(Vectors.dense(4, 1.5), 0), + Row.of(Vectors.dense(4, 4), 1), + Row.of(Vectors.dense(4, 0), 0)); + + private static final List<Row> EUCLIDEAN_WARD_THRESHOLD_AS_TWO_RESULT = + Arrays.asList( + Row.of(Vectors.dense(1, 1), 0), + Row.of(Vectors.dense(1, 4), 1), + Row.of(Vectors.dense(1, 0), 0), + Row.of(Vectors.dense(4, 1.5), 2), + Row.of(Vectors.dense(4, 4), 1), + Row.of(Vectors.dense(4, 0), 2)); + + private static final double TOLERANCE = 1e-7; + + @Before + public void before() { + Configuration config = new Configuration(); + config.set(ExecutionCheckpointingOptions.ENABLE_CHECKPOINTS_AFTER_TASKS_FINISH, true); + env = StreamExecutionEnvironment.getExecutionEnvironment(config); + env.setParallelism(3); + env.enableCheckpointing(100); + env.setRestartStrategy(RestartStrategies.noRestart()); + tEnv = StreamTableEnvironment.create(env); + inputDataTable = tEnv.fromDataStream(env.fromCollection(INPUT_DATA)).as("features"); + } + + @Test + public void testParam() { + AgglomerativeClustering agglomerativeClustering = new AgglomerativeClustering(); + assertEquals("features", agglomerativeClustering.getFeaturesCol()); + assertEquals(2, agglomerativeClustering.getK().intValue()); + assertNull(agglomerativeClustering.getDistanceThreshold()); + assertEquals(AgglomerativeClustering.LINKAGE_WARD, agglomerativeClustering.getLinkage()); + assertEquals(EuclideanDistanceMeasure.NAME, agglomerativeClustering.getDistanceMeasure()); + assertFalse(agglomerativeClustering.getComputeFullTree()); + assertEquals("prediction", agglomerativeClustering.getPredictionCol()); + + agglomerativeClustering + .setFeaturesCol("test_features") + .setK(null) + .setDistanceThreshold(0.01) + .setLinkage(AgglomerativeClusteringParams.LINKAGE_AVERAGE) + .setDistanceMeasure(CosineDistanceMeasure.NAME) + .setComputeFullTree(true) + .setPredictionCol("cluster_id"); + + assertEquals("test_features", agglomerativeClustering.getFeaturesCol()); + assertNull(agglomerativeClustering.getK()); + assertEquals(0.01, agglomerativeClustering.getDistanceThreshold(), TOLERANCE); + assertEquals(AgglomerativeClustering.LINKAGE_AVERAGE, agglomerativeClustering.getLinkage()); + assertEquals(CosineDistanceMeasure.NAME, agglomerativeClustering.getDistanceMeasure()); + assertTrue(agglomerativeClustering.getComputeFullTree()); + assertEquals("cluster_id", agglomerativeClustering.getPredictionCol()); + } + + @Test + public void testOutputSchema() { + Table tempTable = + tEnv.fromDataStream(env.fromElements(Row.of("", ""))) + .as("test_features", "dummy_input"); + AgglomerativeClustering agglomerativeClustering = + new AgglomerativeClustering() + .setFeaturesCol("test_features") + .setPredictionCol("test_prediction"); + Table[] outputs = agglomerativeClustering.transform(tempTable); + assertEquals(2, outputs.length); + assertEquals( + Arrays.asList("test_features", "dummy_input", "test_prediction"), + outputs[0].getResolvedSchema().getColumnNames()); + assertEquals( + Arrays.asList("clusterId1", "clusterId2", "distance", "sizeOfMergedCluster"), + outputs[1].getResolvedSchema().getColumnNames()); + } + + @Test + public void testTransform() throws Exception { + Table[] outputs; + AgglomerativeClustering agglomerativeClustering = + new AgglomerativeClustering() + .setLinkage(AgglomerativeClusteringParams.LINKAGE_AVERAGE) + .setDistanceMeasure(EuclideanDistanceMeasure.NAME) + .setPredictionCol("pred"); + + // Tests euclidean distance with linkage as average, k = 2. + outputs = agglomerativeClustering.transform(inputDataTable); + verifyClusteringResult(EUCLIDEAN_WARD_K_AS_TWO_RESULT, outputs[0]); + + // Tests euclidean distance with linkage as average, k = 2, compute_full_tree = true. + outputs = agglomerativeClustering.setComputeFullTree(true).transform(inputDataTable); + verifyClusteringResult(EUCLIDEAN_WARD_K_AS_TWO_RESULT, outputs[0]); + + // Tests euclidean distance with linkage as average, distance_threshold = 2. + outputs = + agglomerativeClustering + .setK(null) + .setDistanceThreshold(2.0) + .transform(inputDataTable); + verifyClusteringResult(EUCLIDEAN_WARD_THRESHOLD_AS_TWO_RESULT, outputs[0]); + } + + @Test + public void testMergeInfo() throws Exception { + Table[] outputs; + AgglomerativeClustering agglomerativeClustering = + new AgglomerativeClustering() + .setLinkage(AgglomerativeClusteringParams.LINKAGE_AVERAGE) + .setDistanceMeasure(EuclideanDistanceMeasure.NAME) + .setPredictionCol("pred") + .setComputeFullTree(true); + + // Tests euclidean distance with linkage as average. + outputs = agglomerativeClustering.transform(inputDataTable); + verifyMergeInfo(EUCLIDEAN_AVERAGE_MERGE_INFO, outputs[1]); + + // Tests cosine distance with linkage as average. + agglomerativeClustering.setDistanceMeasure(CosineDistanceMeasure.NAME); + outputs = agglomerativeClustering.transform(inputDataTable); + verifyMergeInfo(COSINE_AVERAGE_MERGE_INFO, outputs[1]); + + // Tests manhattan distance with linkage as average. + agglomerativeClustering.setDistanceMeasure(ManhattanDistanceMeasure.NAME); + outputs = agglomerativeClustering.transform(inputDataTable); + verifyMergeInfo(MANHATTAN_AVERAGE_MERGE_INFO, outputs[1]); + + // Tests euclidean distance with linkage as complete. + agglomerativeClustering + .setDistanceMeasure(EuclideanDistanceMeasure.NAME) + .setLinkage(AgglomerativeClusteringParams.LINKAGE_COMPLETE); + outputs = agglomerativeClustering.transform(inputDataTable); + verifyMergeInfo(EUCLIDEAN_COMPLETE_MERGE_INFO, outputs[1]); + + // Tests euclidean distance with linkage as single. + agglomerativeClustering.setLinkage(AgglomerativeClusteringParams.LINKAGE_SINGLE); + outputs = agglomerativeClustering.transform(inputDataTable); + verifyMergeInfo(EUCLIDEAN_SINGLE_MERGE_INFO, outputs[1]); + + // Tests euclidean distance with linkage as ward. + agglomerativeClustering.setLinkage(AgglomerativeClusteringParams.LINKAGE_WARD); + outputs = agglomerativeClustering.transform(inputDataTable); + verifyMergeInfo(EUCLIDEAN_WARD_MERGE_INFO, outputs[1]); + + // Tests merge info not fully computed. + agglomerativeClustering.setComputeFullTree(false); + outputs = agglomerativeClustering.transform(inputDataTable); + verifyMergeInfo(EUCLIDEAN_WARD_MERGE_INFO.subList(0, 4), outputs[1]); + } + + @Test + public void testSaveLoadTransform() throws Exception { + AgglomerativeClustering agglomerativeClustering = + new AgglomerativeClustering() + .setLinkage(AgglomerativeClusteringParams.LINKAGE_AVERAGE) + .setDistanceMeasure(EuclideanDistanceMeasure.NAME) + .setPredictionCol("pred"); + + agglomerativeClustering = + TestUtils.saveAndReload( + tEnv, agglomerativeClustering, tempFolder.newFolder().getAbsolutePath()); + + Table[] outputs = agglomerativeClustering.transform(inputDataTable); + verifyClusteringResult(EUCLIDEAN_WARD_K_AS_TWO_RESULT, outputs[0]); + } + + @SuppressWarnings("unchecked") + private void verifyMergeInfo(List<Row> expected, Table mergeInfoTable) throws Exception { + List<Row> mergeInfo = + IteratorUtils.toList(tEnv.toDataStream(mergeInfoTable).executeAndCollect()); + assertEquals(expected.size(), mergeInfo.size()); + mergeInfo.sort(Comparator.comparingInt(o -> o.getFieldAs(0))); Review Comment: `mergeInfo` should be order-sensitive. I think we should verify that the order of the output merge information is also correct. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
