weibozhao commented on a change in pull request #24: URL: https://github.com/apache/flink-ml/pull/24#discussion_r766444735
########## File path: flink-ml-lib/src/main/java/org/apache/flink/ml/classification/knn/KnnModel.java ########## @@ -0,0 +1,317 @@ +/* + * 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.classification.knn; + +import org.apache.flink.api.common.functions.RichMapFunction; +import org.apache.flink.api.common.typeinfo.BasicTypeInfo; +import org.apache.flink.api.java.tuple.Tuple2; +import org.apache.flink.api.java.tuple.Tuple3; +import org.apache.flink.api.java.typeutils.RowTypeInfo; +import org.apache.flink.ml.api.Model; +import org.apache.flink.ml.classification.knn.KnnModelData.ModelDataDecoder; +import org.apache.flink.ml.common.broadcast.BroadcastUtils; +import org.apache.flink.ml.common.datastream.TableUtils; +import org.apache.flink.ml.linalg.BLAS; +import org.apache.flink.ml.linalg.DenseMatrix; +import org.apache.flink.ml.linalg.DenseVector; +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.streaming.api.environment.StreamExecutionEnvironment; +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.commons.lang3.ArrayUtils; + +import java.io.IOException; +import java.util.ArrayList; +import java.util.Collections; +import java.util.Comparator; +import java.util.HashMap; +import java.util.List; +import java.util.Map; +import java.util.PriorityQueue; + +/** Knn model fitted by estimator. */ +public class KnnModel implements Model<KnnModel>, KnnModelParams<KnnModel> { + protected Map<Param<?>, Object> params = new HashMap<>(); + private Table modelDataTable; + + public KnnModel() { + ParamUtils.initializeMapWithDefaultValues(params, this); + } + + @Override + public KnnModel setModelData(Table... modelData) { + this.modelDataTable = modelData[0]; + return this; + } + + @Override + public Table[] getModelData() { + return new Table[] {modelDataTable}; + } + + @Override + @SuppressWarnings("unchecked") + public Table[] transform(Table... inputs) { + StreamTableEnvironment tEnv = + (StreamTableEnvironment) ((TableImpl) inputs[0]).getTableEnvironment(); + DataStream<Row> input = tEnv.toDataStream(inputs[0]); + DataStream<KnnModelData> model = KnnModelData.getModelDataStream(modelDataTable); + final String broadcastKey = "broadcastModelKey"; + + RowTypeInfo inputTypeInfo = TableUtils.getRowTypeInfo(inputs[0].getResolvedSchema()); + RowTypeInfo outputTypeInfo = + new RowTypeInfo( + ArrayUtils.addAll( + inputTypeInfo.getFieldTypes(), BasicTypeInfo.DOUBLE_TYPE_INFO), + ArrayUtils.addAll(inputTypeInfo.getFieldNames(), getPredictionCol())); + + DataStream<Row> output = + BroadcastUtils.withBroadcastStream( + Collections.singletonList(input), + Collections.singletonMap(broadcastKey, model), + inputList -> { + DataStream inoutData = inputList.get(0); + return inoutData.map( + new PredictLabelFunction( + broadcastKey, getK(), getFeaturesCol()), + outputTypeInfo); + }); + + return new Table[] {tEnv.fromDataStream(output)}; + } + + @Override + public Map<Param<?>, Object> getParamMap() { + return this.params; + } + + @Override + public void save(String path) throws IOException { + ReadWriteUtils.saveMetadata(this, path); + ReadWriteUtils.saveModelData( + KnnModelData.getModelDataStream(modelDataTable), + path, + new KnnModelData.ModelDataEncoder(), + 0); + } + + /** + * Loads model data from path. + * + * @param env Stream execution environment. + * @param path Model path. + * @return Knn model. + */ + public static KnnModel load(StreamExecutionEnvironment env, String path) throws IOException { + KnnModel model = ReadWriteUtils.loadStageParam(path); + Table modelDataTable = ReadWriteUtils.loadModelData(env, path, new ModelDataDecoder(), 0); + return model.setModelData(modelDataTable); + } + + /** This operator loads model data and predicts result. */ + private static class PredictLabelFunction extends RichMapFunction<Row, Row> { + private boolean firstElement = true; + private final String featureCol; + private transient KnnModelData knnModelData; + private final int topN; + private final String broadcastKey; + private transient Comparator<? super Tuple2<Double, Double>> comparator; + + public PredictLabelFunction(String broadcastKey, int k, String featureCol) { + this.topN = k; + this.broadcastKey = broadcastKey; + this.featureCol = featureCol; + } + + @Override + public Row map(Row row) { + if (firstElement) { + comparator = Comparator.comparingDouble(o -> -o.f0); + loadModel(getRuntimeContext().getBroadcastVariable(broadcastKey)); + firstElement = false; + } + DenseVector vector = (DenseVector) row.getField(featureCol); + Tuple2<List<Double>, List<Double>> t2 = findNeighbor(vector, topN); + return Row.join(row, Row.of(getResult(t2))); + } + + /** + * Finds the nearest topN neighbors from whole nodes. + * + * @param input Input vector. + * @param topN Top N. + * @return Neighbors. + */ + private Tuple2<List<Double>, List<Double>> findNeighbor(DenseVector input, Integer topN) { + PriorityQueue<Tuple2<Double, Double>> priorityQueue = new PriorityQueue<>(comparator); + search(input, topN, priorityQueue); + List<Double> items = new ArrayList<>(); + List<Double> metrics = new ArrayList<>(); + while (!priorityQueue.isEmpty()) { + Tuple2<Double, Double> result = priorityQueue.poll(); + items.add(result.f1); + metrics.add(result.f0); + } + Collections.reverse(items); + Collections.reverse(metrics); + priorityQueue.clear(); Review comment: OK, I will refine it. -- 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: issues-unsubscr...@flink.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
