zhipeng93 commented on code in PR #90: URL: https://github.com/apache/flink-ml/pull/90#discussion_r858415540
########## flink-ml-lib/src/main/java/org/apache/flink/ml/common/glm/LocalTrainer.java: ########## @@ -0,0 +1,177 @@ +/* + * 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.common.glm; + +import org.apache.flink.api.common.state.ListState; +import org.apache.flink.api.common.state.ListStateDescriptor; +import org.apache.flink.api.common.typeinfo.PrimitiveArrayTypeInfo; +import org.apache.flink.api.common.typeinfo.TypeInformation; +import org.apache.flink.iteration.operator.OperatorStateUtils; +import org.apache.flink.ml.linalg.DenseVector; +import org.apache.flink.ml.regression.linearregression.LinearRegression; +import org.apache.flink.runtime.state.FunctionInitializationContext; +import org.apache.flink.runtime.state.FunctionSnapshotContext; +import org.apache.flink.streaming.api.checkpoint.CheckpointedFunction; + +import java.util.Iterator; + +/** + * A local trainer is a trainer that uses a batch of training data to compute a update of the model + * locally. + * + * @param <T> Class type of training data. + */ +public abstract class LocalTrainer<T> implements CheckpointedFunction { Review Comment: Thanks for the insightful comments. > move CacheDataAndDoTrain to an independent class It is an dependent class in the old PR. Do you mean to merge it with LocalTrainer and WithRegularization? > merge LocalTrainer and WithRegularization and remove methods like getReg, as now they are only used internally It seems hard to merge these two class, because the logic of `WithRegularization` is supposed to be used in `LocalTrainer#updateModel`. I have renamed `WithRegularization` as `RegularizationUtils` and removed methods like `withReg`. What do you think? > rename the merged class. A name like LocalTrainer might be too general to be associated with linear algorithms. I have renamed the class to `LocalLinearTrainer`. I still think it is not a very good name. We could probably discuss more on the naming. > merge CacheDataAndDoTrain, LocalTrainer and WithRegularization. I did not merge `CacheDataAndDoTrain` with the other two for now for the following reasons: - `CacheDataAndDoTrain` is a more a infra and involves distributed concepts, like two input operators. - `LocalTrainer` is a more friendly and clean concept for machine learning users, since it only involves local operations. > change the API and implementation of trainOnBatchData. I have the sense that each time only one data, instead of a batch of data, is enough. I think the API of org.apache.flink.api.common.functions.AggregateFunction is a good reference. I did not do the change for the following reasons: - mini-batch training is a common concept for machine learning. - Users may want to do operations before/after mini-batch training. ########## flink-ml-lib/src/main/java/org/apache/flink/ml/classification/logisticregression/LogisticRegressionModelData.java: ########## @@ -42,12 +43,9 @@ * <p>This class also provides methods to convert model data from Table to Datastream, and classes * to save/load model data. */ -public class LogisticRegressionModelData { - - public DenseVector coefficient; - +public class LogisticRegressionModelData extends GeneralLinearAlgoModelData { Review Comment: I also tried to do this but failed because in `ModelDataEncoder` we need to construct an instance of `LinearRegressionModelData` and `LogisticRegressionModelData `. If we pass a class type, we may need to go with reflections, which is usually not encouraged. ########## flink-ml-lib/src/main/java/org/apache/flink/ml/classification/logisticregression/LogisticRegressionModel.java: ########## @@ -18,51 +18,28 @@ package org.apache.flink.ml.classification.logisticregression; -import org.apache.flink.api.common.functions.RichMapFunction; import org.apache.flink.api.common.typeinfo.BasicTypeInfo; import org.apache.flink.api.common.typeinfo.TypeInformation; -import org.apache.flink.api.java.tuple.Tuple2; 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.common.glm.GeneralLinearAlgoModel; import org.apache.flink.ml.linalg.BLAS; import org.apache.flink.ml.linalg.DenseVector; import org.apache.flink.ml.linalg.Vectors; -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.Collections; -import java.util.HashMap; -import java.util.Map; +import java.io.Serializable; -/** A Model which classifies data using the model data computed by {@link LogisticRegression}. */ -public class LogisticRegressionModel - implements Model<LogisticRegressionModel>, - LogisticRegressionModelParams<LogisticRegressionModel> { - - private final Map<Param<?>, Object> paramMap = new HashMap<>(); - - private Table modelDataTable; - - public LogisticRegressionModel() { - ParamUtils.initializeMapWithDefaultValues(paramMap, this); - } - - @Override - public Map<Param<?>, Object> getParamMap() { - return paramMap; - } +/** + * A Model which classifies the input using the model data computed by {@link LogisticRegression}. + */ +public class LogisticRegressionModel extends GeneralLinearAlgoModel<LogisticRegressionModel> + implements LogisticRegressionModelParams<LogisticRegressionModel>, Serializable { @Override public void save(String path) throws IOException { Review Comment: The change seems infeasible for now... -- 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]
