zhengruifeng commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component URL: https://github.com/apache/spark/pull/26124#discussion_r341439843
########## File path: mllib/src/main/scala/org/apache/spark/ml/regression/FMRegressor.scala ########## @@ -0,0 +1,768 @@ +/* + * 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.spark.ml.regression + +import scala.util.Random + +import breeze.linalg.{axpy => brzAxpy, norm => brzNorm, Vector => BV} +import breeze.numerics.{sqrt => brzSqrt} +import org.apache.hadoop.fs.Path + +import org.apache.spark.annotation.Since +import org.apache.spark.internal.Logging +import org.apache.spark.ml.{PredictionModel, Predictor, PredictorParams} +import org.apache.spark.ml.linalg._ +import org.apache.spark.ml.linalg.BLAS._ +import org.apache.spark.ml.param._ +import org.apache.spark.ml.param.shared._ +import org.apache.spark.ml.util._ +import org.apache.spark.ml.util.Instrumentation.instrumented +import org.apache.spark.mllib.{linalg => OldLinalg} +import org.apache.spark.mllib.linalg.{Vector => OldVector, Vectors => OldVectors} +import org.apache.spark.mllib.linalg.VectorImplicits._ +import org.apache.spark.mllib.optimization.{Gradient, GradientDescent, SquaredL2Updater, Updater} +import org.apache.spark.mllib.regression.{LabeledPoint => OldLabeledPoint} +import org.apache.spark.mllib.util.MLUtils +import org.apache.spark.rdd.RDD +import org.apache.spark.sql.{Dataset, Row} +import org.apache.spark.sql.functions.col +import org.apache.spark.storage.StorageLevel + +/** + * Params for Factorization Machines + */ +private[ml] trait FactorizationMachinesParams + extends PredictorParams + with HasMaxIter with HasStepSize with HasTol with HasSolver { + + /** + * Param for dimensionality of the factors (>= 0) + * @group param + */ + @Since("3.0.0") + final val numFactors: IntParam = new IntParam(this, "numFactors", + "Dimensionality of the factor vectors, " + + "which are used to get pairwise interactions between variables", + ParamValidators.gt(0)) + + /** @group getParam */ + @Since("3.0.0") + final def getNumFactors: Int = $(numFactors) + + /** + * Param for whether to fit global bias term + * @group param + */ + @Since("3.0.0") + final val fitBias: BooleanParam = new BooleanParam(this, "fitBias", + "whether to fit global bias term") + + /** @group getParam */ + @Since("3.0.0") + final def getFitBias: Boolean = $(fitBias) + + /** + * Param for whether to fit linear term (aka 1-way term) + * @group param + */ + @Since("3.0.0") + final val fitLinear: BooleanParam = new BooleanParam(this, "fitLinear", + "whether to fit linear term (aka 1-way term)") + + /** @group getParam */ + @Since("3.0.0") + final def getFitLinear: Boolean = $(fitLinear) + + /** + * Param for L2 regularization parameter (>= 0) + * @group param + */ + @Since("3.0.0") + final val regParam: DoubleParam = new DoubleParam(this, "regParam", + "the magnitude of L2-regularization", ParamValidators.gtEq(0)) + + /** @group getParam */ + @Since("3.0.0") + final def getRegParam: Double = $(regParam) + + /** + * Param for mini-batch fraction, must be in range (0, 1] + * @group param + */ + @Since("3.0.0") + final val miniBatchFraction: DoubleParam = new DoubleParam(this, "miniBatchFraction", + "fraction of the input data set that should be used for one iteration of gradient descent", + ParamValidators.inRange(0, 1, false, true)) + + /** @group getParam */ + @Since("3.0.0") + final def getMiniBatchFraction: Double = $(miniBatchFraction) + + /** + * Param for standard deviation of initial coefficients + * @group param + */ + @Since("3.0.0") + final val initStd: DoubleParam = new DoubleParam(this, "initStd", + "standard deviation of initial coefficients", ParamValidators.gt(0)) + + /** @group getParam */ + @Since("3.0.0") + final def getInitStd: Double = $(initStd) + + /** String name for "gd". */ + private[ml] val GD = "gd" + + /** String name for "adamW". */ + private[ml] val AdamW = "adamW" + + /** Set of solvers that FactorizationMachines supports. */ + private[ml] val supportedSolvers = Array(GD, AdamW) + + /** + * The solver algorithm for optimization. + * Supported options: "gd", "adamW". + * Default: "adamW" + * + * @group param + */ + @Since("3.0.0") + final override val solver: Param[String] = new Param[String](this, "solver", + "The solver algorithm for optimization. Supported options: " + + s"${supportedSolvers.mkString(", ")}. (Default adamW)", + ParamValidators.inArray[String](supportedSolvers)) + + private[ml] def parseSolver(solver: String, coefficientsSize: Int): Updater = { + solver match { + case GD => new SquaredL2Updater() + case AdamW => new AdamWUpdater(coefficientsSize) + } + } +} + +/** + * Params for FMRegressor + */ +private[regression] trait FMRegressorParams extends FactorizationMachinesParams { +} + +/** + * Factorization Machines learning algorithm for regression. + * It supports normal gradient descent and AdamW solver. + * + * The implementation is based upon: + * <a href="https://www.csie.ntu.edu.tw/~b97053/paper/Rendle2010FM.pdf";> + * S. Rendle. "Factorization machines" 2010</a>. + * + * FM is able to estimate interactions even in problems with huge sparsity + * (like advertising and recommendation system). + * FM formula is: + * {{{ + * y = w_0 + \sum\limits^n_{i-1} w_i x_i + + * \sum\limits^n_{i=1} \sum\limits^n_{j=i+1} \langle v_i, v_j \rangle x_i x_j + * }}} + * First two terms denote global bias and linear term (as same as linear regression), + * and last term denotes pairwise interactions term. {{{v_i}}} describes the i-th variable + * with k factors. + * + * FM regression model uses MSE loss which can be solved by gradient descent method, and + * regularization terms like L2 are usually added to the loss function to prevent overfitting. + */ +@Since("3.0.0") +class FMRegressor @Since("3.0.0") ( + @Since("3.0.0") override val uid: String) + extends Predictor[Vector, FMRegressor, FMRegressorModel] + with FMRegressorParams with DefaultParamsWritable with Logging { + + import org.apache.spark.ml.regression.BaseFactorizationMachinesGradient.{SquaredError, parseLoss} + import org.apache.spark.ml.regression.FMRegressor.initCoefficients + + @Since("3.0.0") + def this() = this(Identifiable.randomUID("fmr")) + + /** + * Set the dimensionality of the factors. + * Default is 8. + * + * @group setParam + */ + @Since("3.0.0") + def setNumFactors(value: Int): this.type = set(numFactors, value) + setDefault(numFactors -> 8) + + /** + * Set whether to fit global bias term. + * Default is true. + * + * @group setParam + */ + @Since("3.0.0") + def setFitBias(value: Boolean): this.type = set(fitBias, value) + setDefault(fitBias -> true) + + /** + * Set whether to fit linear term. + * Default is true. + * + * @group setParam + */ + @Since("3.0.0") + def setFitLinear(value: Boolean): this.type = set(fitLinear, value) + setDefault(fitLinear -> true) + + /** + * Set the L2 regularization parameter. + * Default is 0.0. + * + * @group setParam + */ + @Since("3.0.0") + def setRegParam(value: Double): this.type = set(regParam, value) + setDefault(regParam -> 0.0) + + /** + * Set the mini-batch fraction parameter. + * Default is 1.0. + * + * @group setParam + */ + @Since("3.0.0") + def setMiniBatchFraction(value: Double): this.type = { + require(value > 0 && value <= 1.0, Review comment: `ParamValidators.inRange(0, 1, false, true)` already checks input value ---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
