lindong28 commented on code in PR #187: URL: https://github.com/apache/flink-ml/pull/187#discussion_r1044092761
########## flink-ml-lib/src/main/java/org/apache/flink/ml/feature/univariatefeatureselector/UnivariateFeatureSelector.java: ########## @@ -0,0 +1,305 @@ +/* + * 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.feature.univariatefeatureselector; + +import org.apache.flink.api.common.state.ListState; +import org.apache.flink.api.common.state.ListStateDescriptor; +import org.apache.flink.api.common.typeinfo.TypeInformation; +import org.apache.flink.api.common.typeinfo.Types; +import org.apache.flink.api.java.tuple.Tuple2; +import org.apache.flink.ml.api.Estimator; +import org.apache.flink.ml.param.Param; +import org.apache.flink.ml.stats.anovatest.ANOVATest; +import org.apache.flink.ml.stats.chisqtest.ChiSqTest; +import org.apache.flink.ml.stats.fvaluetest.FValueTest; +import org.apache.flink.ml.util.ParamUtils; +import org.apache.flink.ml.util.ReadWriteUtils; +import org.apache.flink.runtime.state.StateInitializationContext; +import org.apache.flink.runtime.state.StateSnapshotContext; +import org.apache.flink.streaming.api.datastream.DataStream; +import org.apache.flink.streaming.api.operators.AbstractStreamOperator; +import org.apache.flink.streaming.api.operators.BoundedOneInput; +import org.apache.flink.streaming.api.operators.OneInputStreamOperator; +import org.apache.flink.streaming.runtime.streamrecord.StreamRecord; +import org.apache.flink.table.api.Table; +import org.apache.flink.table.api.bridge.java.StreamTableEnvironment; +import org.apache.flink.table.api.internal.TableImpl; +import org.apache.flink.types.Row; +import org.apache.flink.util.Preconditions; + +import org.apache.commons.collections.IteratorUtils; + +import java.io.IOException; +import java.util.ArrayList; +import java.util.Comparator; +import java.util.HashMap; +import java.util.List; +import java.util.Map; +import java.util.stream.IntStream; + +/** + * An Estimator which selects features based on univariate statistical tests against labels. + * + * <p>Currently, Flink supports three Univariate Feature Selectors: chi-squared, ANOVA F-test and + * F-value. User can choose Univariate Feature Selector by setting `featureType` and `labelType`, + * and Flink will pick the score function based on the specified `featureType` and `labelType`. + * + * <p>The following combination of `featureType` and `labelType` are supported: + * + * <ul> + * <li>`featureType` `categorical` and `labelType` `categorical`: Flink uses chi-squared, i.e. + * chi2 in sklearn. + * <li>`featureType` `continuous` and `labelType` `categorical`: Flink uses ANOVA F-test, + * f_classif in sklearn. Review Comment: nits: add `i.e.` for consistency. ########## flink-ml-lib/src/main/java/org/apache/flink/ml/feature/univariatefeatureselector/UnivariateFeatureSelectorModel.java: ########## @@ -0,0 +1,201 @@ +/* + * 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.feature.univariatefeatureselector; + +import org.apache.flink.api.common.functions.RichMapFunction; +import org.apache.flink.api.java.typeutils.RowTypeInfo; +import org.apache.flink.ml.api.Model; +import org.apache.flink.ml.common.broadcast.BroadcastUtils; +import org.apache.flink.ml.common.datastream.TableUtils; +import org.apache.flink.ml.linalg.DenseVector; +import org.apache.flink.ml.linalg.SparseVector; +import org.apache.flink.ml.linalg.Vector; +import org.apache.flink.ml.linalg.Vectors; +import org.apache.flink.ml.linalg.typeinfo.VectorTypeInfo; +import org.apache.flink.ml.param.Param; +import org.apache.flink.ml.util.ParamUtils; +import org.apache.flink.ml.util.ReadWriteUtils; +import org.apache.flink.streaming.api.datastream.DataStream; +import org.apache.flink.table.api.Table; +import org.apache.flink.table.api.bridge.java.StreamTableEnvironment; +import org.apache.flink.table.api.internal.TableImpl; +import org.apache.flink.types.Row; +import org.apache.flink.util.Preconditions; + +import org.apache.commons.lang3.ArrayUtils; + +import java.io.IOException; +import java.util.ArrayList; +import java.util.Arrays; +import java.util.Collections; +import java.util.HashMap; +import java.util.List; +import java.util.Map; + +/** + * A Model which transforms data using the model data computed by {@link UnivariateFeatureSelector}. + */ +public class UnivariateFeatureSelectorModel + implements Model<UnivariateFeatureSelectorModel>, + UnivariateFeatureSelectorModelParams<UnivariateFeatureSelectorModel> { + + private final Map<Param<?>, Object> paramMap = new HashMap<>(); + private Table modelDataTable; + + public UnivariateFeatureSelectorModel() { + ParamUtils.initializeMapWithDefaultValues(paramMap, this); + } + + @Override + public UnivariateFeatureSelectorModel setModelData(Table... inputs) { + Preconditions.checkArgument(inputs.length == 1); + modelDataTable = inputs[0]; + return this; + } + + @Override + public Table[] getModelData() { + return new Table[] {modelDataTable}; + } + + @SuppressWarnings({"unchecked", "rawtypes"}) + @Override + public Table[] transform(Table... inputs) { + Preconditions.checkArgument(inputs.length == 1); + + StreamTableEnvironment tEnv = + (StreamTableEnvironment) ((TableImpl) inputs[0]).getTableEnvironment(); + DataStream<Row> data = tEnv.toDataStream(inputs[0]); + DataStream<UnivariateFeatureSelectorModelData> modelData = + UnivariateFeatureSelectorModelData.getModelDataStream(modelDataTable); + + final String broadcastModelKey = "broadcastModelKey"; + RowTypeInfo inputTypeInfo = TableUtils.getRowTypeInfo(inputs[0].getResolvedSchema()); + RowTypeInfo outputTypeInfo = + new RowTypeInfo( + ArrayUtils.addAll(inputTypeInfo.getFieldTypes(), VectorTypeInfo.INSTANCE), + ArrayUtils.addAll(inputTypeInfo.getFieldNames(), getOutputCol())); + + DataStream<Row> outputStream = + BroadcastUtils.withBroadcastStream( + Collections.singletonList(data), + Collections.singletonMap(broadcastModelKey, modelData), + inputList -> { + DataStream input = inputList.get(0); + return input.map( + new PredictOutputFunction(getFeaturesCol(), broadcastModelKey), + outputTypeInfo); + }); + + return new Table[] {tEnv.fromDataStream(outputStream)}; + } + + /** This operator loads model data and predicts result. */ + private static class PredictOutputFunction extends RichMapFunction<Row, Row> { + + private final String inputCol; + private final String broadcastKey; + private int[] indices; + + public PredictOutputFunction(String inputCol, String broadcastKey) { + this.inputCol = inputCol; + this.broadcastKey = broadcastKey; + } + + @Override + public Row map(Row row) { + if (indices == null) { + UnivariateFeatureSelectorModelData modelData = + (UnivariateFeatureSelectorModelData) + getRuntimeContext().getBroadcastVariable(broadcastKey).get(0); + indices = Arrays.stream(modelData.indices).sorted().toArray(); + } + + if (indices.length == 0) { + return Row.join(row, Row.of(Vectors.dense())); + } else { + Vector inputVec = ((Vector) row.getField(inputCol)); + Preconditions.checkArgument( + inputVec.size() > indices[indices.length - 1], + "Input %s features, but UnivariateFeatureSelector is " + + "expecting at least %s features as input.", + inputVec.size(), + indices[indices.length - 1] + 1); + Vector outputVec = selectByIndices(inputVec, indices); + return Row.join(row, Row.of(outputVec)); + } + } + + /** + * Selects a subset of the vector base on the indices. Note that the input indices must be + * sorted in ascending order if the input vector is sparse. + */ + private Vector selectByIndices(Vector vector, int[] selectedIndices) { Review Comment: Since this method looks pretty generic and it is re-used by `VarianceThresholdSelectorModel.java`, how about we move this method to `org.apache.flink.ml.common.util.VectorUtils` and make it static? -- 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]
