[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r356386089
##
File path:
mllib/src/test/scala/org/apache/spark/ml/classification/FMClassifierSuite.scala
##
@@ -0,0 +1,143 @@
+/*
+ * 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.classification
+
+import
org.apache.spark.ml.classification.LogisticRegressionSuite.generateLogisticInput
+import org.apache.spark.ml.linalg.{DenseVector, Vector, Vectors}
+import org.apache.spark.ml.param.ParamsSuite
+import org.apache.spark.ml.regression.FMRegressorSuite._
+import org.apache.spark.ml.util._
+import org.apache.spark.ml.util.TestingUtils._
+import org.apache.spark.sql.{DataFrame, Row}
+
+class FMClassifierSuite extends MLTest with DefaultReadWriteTest {
Review comment:
> Is this resolved? I'd also love to have at least one test that compares
output to some other library, even if the test can only assert some approximate
match in loss, etc.
Now, test method is FM fit a dataset perfectly (loss closed to zero and
model's weights closed to real weights), the dataset is generated by a group
weight (include bias/linear/factor weights). I think it is enough.
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:
us...@infra.apache.org
With regards,
Apache Git Services
-
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r356381863
##
File path:
mllib/src/main/scala/org/apache/spark/ml/regression/FMRegressor.scala
##
@@ -0,0 +1,839 @@
+/*
+ * 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.regression.FactorizationMachines._
+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.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 with HasSeed {
+
+ /**
+ * Param for dimensionality of the factors (= 0)
+ * @group param
+ */
+ @Since("3.0.0")
+ final val factorSize: IntParam = new IntParam(this, "factorSize",
+"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 getFactorSize: Int = $(factorSize)
+
+ /**
+ * 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)
+
+ /**
+ * The solver algorithm for optimization.
+ * Supported options: "gd", "adamW".
+ * Default: "adamW"
+ *
+ * @group param
+ */
+ @Since("3.0.0")
+ final override val solver: Param[String] = new
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r356381039
##
File path:
mllib/src/main/scala/org/apache/spark/ml/regression/FMRegressor.scala
##
@@ -0,0 +1,839 @@
+/*
+ * 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.regression.FactorizationMachines._
+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.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 with HasSeed {
+
+ /**
+ * Param for dimensionality of the factors (= 0)
+ * @group param
+ */
+ @Since("3.0.0")
+ final val factorSize: IntParam = new IntParam(this, "factorSize",
+"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 getFactorSize: Int = $(factorSize)
+
+ /**
+ * 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)
+
+ /**
+ * The solver algorithm for optimization.
+ * Supported options: "gd", "adamW".
+ * Default: "adamW"
+ *
+ * @group param
+ */
+ @Since("3.0.0")
+ final override val solver: Param[String] = new
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r356380671
##
File path: python/pyspark/ml/regression.py
##
@@ -2126,6 +2127,172 @@ def __repr__(self):
return self._call_java("toString")
+@inherit_doc
+class FactorizationMachines(JavaPredictor, HasMaxIter, HasStepSize, HasTol,
HasSolver, HasLoss,
+JavaMLWritable, JavaMLReadable):
+"""
+Factorization Machines.
+
+loss Supports:
+
+* logisticLoss (default)
+* squaredError
+
+solver Supports:
+
+* gd (normal mini-batch gradient descent)
+* adamW (default)
+
+>>> from pyspark.ml.linalg import Vectors
+>>> from pyspark.ml.regression import FactorizationMachines
+>>> df = spark.createDataFrame([
+... (1.0, Vectors.dense(1.0)),
+... (0.0, Vectors.sparse(1, [], []))], ["label", "features"])
+>>> fm = FactorizationMachines(numFactors=2, )
+>>> model = fm.fit(df)
+>>> test0 = spark.createDataFrame([
+... (Vectors.dense(-1.0),),
+... (Vectors.dense(0.5),),
+... (Vectors.dense(1.0),),
+... (Vectors.dense(2.0),)], ["features"])
+>>> model.transform(test0).show(10, False)
+++--+
+|features|prediction|
+++--+
+|[-1.0] |1.7219239347644947E-10|
+|[0.5] |0.612429917023823 |
+|[1.0] |0.99969782036162 |
+|[2.0] |0.99310394|
+++--+
+
+.. versionadded:: 3.0.0
+"""
+
+numFactors = Param(Params._dummy(), "numFactors", "dimensionality of the
factor vectors, " +
+ "which are used to get pairwise interactions between
variables",
Review comment:
It is resolved. The doc strings should start with a Capital (dimensionality
of .. -> Dimensionality of ...).
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:
us...@infra.apache.org
With regards,
Apache Git Services
-
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r355913166
##
File path:
mllib/src/main/scala/org/apache/spark/ml/regression/FMRegressor.scala
##
@@ -0,0 +1,826 @@
+/*
+ * 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.regression.FactorizationMachines._
+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.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 with HasSeed {
+
+ /**
+ * Param for dimensionality of the factors (= 0)
+ * @group param
+ */
+ @Since("3.0.0")
+ final val factorSize: IntParam = new IntParam(this, "factorSize",
+"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 getFactorSize: Int = $(factorSize)
+
+ /**
+ * 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)
+
+ /**
+ * The solver algorithm for optimization.
+ * Supported options: "gd", "adamW".
+ * Default: "adamW"
+ *
+ * @group param
+ */
+ @Since("3.0.0")
+ final override val solver: Param[String] = new
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r355938752
##
File path:
mllib/src/test/scala/org/apache/spark/ml/regression/FMRegressorSuite.scala
##
@@ -0,0 +1,240 @@
+/*
+ * 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 org.apache.spark.ml.linalg._
+import org.apache.spark.ml.param.ParamsSuite
+import org.apache.spark.ml.regression.FactorizationMachines._
+import org.apache.spark.ml.regression.FMRegressorSuite._
+import org.apache.spark.ml.util._
+import org.apache.spark.ml.util.TestingUtils._
+import org.apache.spark.sql.{DataFrame, SparkSession}
+import org.apache.spark.sql.functions.{avg, col}
+
+class FMRegressorSuite extends MLTest with DefaultReadWriteTest {
+
+ private val seed = 10
+ @transient var crossDataset: DataFrame = _
+
+ override def beforeAll(): Unit = {
+super.beforeAll()
+
+val (crossDatasetTmp, _) = generateFactorInteractionInput(
+ spark, 2, 10, 1000, seed, true, true)
+crossDataset = crossDatasetTmp
+ }
+
+ test("params") {
+ParamsSuite.checkParams(new FMRegressor)
+val model = new FMRegressionModel("fmr_test", 0.0, Vectors.dense(0.0),
+ new DenseMatrix(1, 8, new Array[Double](8)))
+ParamsSuite.checkParams(model)
+ }
+
+ test("combineCoefficients") {
+val numFeatures = 2
+val factorSize = 4
+val b = 0.1
+val w = Vectors.dense(Array(0.2, 0.3))
+val v = new DenseMatrix(numFeatures, factorSize,
+ Array(0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0, 1.1), true)
+
+val expectList = Array(
+ (true, true, Array(0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0, 1.1, 0.2, 0.3,
0.1)),
+ (false, true, Array(0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0, 1.1, 0.2, 0.3)),
+ (true, false, Array(0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0, 1.1, 0.1)),
+ (false, false, Array(0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0, 1.1)))
+
+expectList.foreach { case (fitBias, fitLinear, expectCoeffs) =>
+ assert(combineCoefficients(b, w, v, fitBias, fitLinear) ===
Vectors.dense(expectCoeffs))
+}
+ }
+
+ test("splitCoefficients") {
+val numFeatures = 2
+val factorSize = 4
+val b = 0.1
+val w = Vectors.dense(Array(0.2, 0.3))
+val v = new DenseMatrix(numFeatures, factorSize,
+ Array(0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0, 1.1), true)
+val emptyB = 0.0
+val emptyW = Vectors.sparse(numFeatures, Seq.empty)
+
+val expectList = Array(
+ (true, true, b, w, v, Array(0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0, 1.1, 0.2,
0.3, 0.1)),
+ (false, true, emptyB, w, v, Array(0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0,
1.1, 0.2, 0.3)),
+ (true, false, b, emptyW, v, Array(0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0,
1.1, 0.1)),
+ (false, false, emptyB, emptyW, v, Array(0.4, 0.5, 0.6, 0.7, 0.8, 0.9,
1.0, 1.1)))
+
+expectList.foreach { case (fitBias, fitLinear, b1, w1, v1, coeffs) =>
+ val (b2, w2, v2) = splitCoefficients(Vectors.dense(coeffs),
+numFeatures, factorSize, fitBias, fitLinear)
+ assert(b1 === b2)
+ assert(w1 === w2)
+ assert(v1 === v2)
+}
+ }
+
+ def checkMSE(fitBias: Boolean, fitLinear: Boolean): Unit = {
+val numFeatures = 10
+val numSamples = 1000
+val factorSize = 4
+val (data, coefficients) = generateFactorInteractionInput(
+ spark, factorSize, numFeatures, numSamples, seed, fitBias, fitLinear)
+val (b, w, v) = splitCoefficients(new DenseVector(coefficients),
+ numFeatures, factorSize, fitBias, fitLinear)
+
+val fm = new FMRegressor()
+ .setSolver("adamW")
+ .setFeaturesCol("features")
+ .setLabelCol("label")
+ .setFactorSize(factorSize)
+ .setFitBias(fitBias)
+ .setFitLinear(fitLinear)
+ .setInitStd(0.01)
+ .setMaxIter(300)
+ .setMiniBatchFraction(1.0)
+ .setStepSize(1.0)
+ .setRegParam(0.0)
+ .setTol(1E-6)
+val fmModel = fm.fit(data)
+val res = fmModel.transform(data)
+
+// check mse value
+val mse = res.select((col("prediction") - col("label")).as("error"))
+ .select((col("error") *
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r355913166
##
File path:
mllib/src/main/scala/org/apache/spark/ml/regression/FMRegressor.scala
##
@@ -0,0 +1,826 @@
+/*
+ * 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.regression.FactorizationMachines._
+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.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 with HasSeed {
+
+ /**
+ * Param for dimensionality of the factors (= 0)
+ * @group param
+ */
+ @Since("3.0.0")
+ final val factorSize: IntParam = new IntParam(this, "factorSize",
+"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 getFactorSize: Int = $(factorSize)
+
+ /**
+ * 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)
+
+ /**
+ * The solver algorithm for optimization.
+ * Supported options: "gd", "adamW".
+ * Default: "adamW"
+ *
+ * @group param
+ */
+ @Since("3.0.0")
+ final override val solver: Param[String] = new
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r344549552
##
File path:
mllib/src/main/scala/org/apache/spark/ml/regression/FactorizationMachines.scala
##
@@ -0,0 +1,757 @@
+/*
+ * 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[regression] trait FactorizationMachinesParams
+ extends PredictorParams
+ with HasMaxIter with HasStepSize with HasTol with HasSolver with HasLoss {
+
+ import FactorizationMachines._
+
+ /**
+ * 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 parameter of l2-regularization term, " +
+ "which prevents overfitting by adding sum of squares of all the
parameters",
+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)
+
+ /**
+ * The solver algorithm for optimization.
+ * Supported
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r344547494
##
File path:
mllib/src/main/scala/org/apache/spark/ml/classification/FMClassifier.scala
##
@@ -0,0 +1,326 @@
+/*
+ * 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.classification
+
+import org.apache.hadoop.fs.Path
+
+import org.apache.spark.annotation.Since
+import org.apache.spark.internal.Logging
+import org.apache.spark.ml.linalg._
+import org.apache.spark.ml.param._
+import org.apache.spark.ml.regression.{FactorizationMachines,
FactorizationMachinesParams}
+import org.apache.spark.ml.regression.FactorizationMachines._
+import org.apache.spark.ml.util._
+import org.apache.spark.ml.util.Instrumentation.instrumented
+import org.apache.spark.mllib.linalg.{Vector => OldVector}
+import org.apache.spark.mllib.linalg.VectorImplicits._
+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 FMClassifier.
+ */
+private[classification] trait FMClassifierParams extends
ProbabilisticClassifierParams
+ with FactorizationMachinesParams {
+}
+
+/**
+ * Factorization Machines learning algorithm for classification.
+ * It supports normal gradient descent and AdamW solver.
+ *
+ * The implementation is based upon:
+ * https://www.csie.ntu.edu.tw/~b97053/paper/Rendle2010FM.pdf;>
+ * S. Rendle. "Factorization machines" 2010.
+ *
+ * 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 classification model uses logistic loss which can be solved by gradient
descent method, and
+ * regularization terms like L2 are usually added to the loss function to
prevent overfitting.
+ *
+ * @note Multiclass labels are not currently supported.
+ */
+@Since("3.0.0")
+class FMClassifier @Since("3.0.0") (
+@Since("3.0.0") override val uid: String)
+ extends ProbabilisticClassifier[Vector, FMClassifier, FMClassifierModel]
+ with FactorizationMachines with FMClassifierParams with
DefaultParamsWritable with Logging {
+
+ @Since("3.0.0")
+ def this() = this(Identifiable.randomUID("fmc"))
+
+ /**
+ * 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 = set(miniBatchFraction,
value)
+ setDefault(miniBatchFraction -> 1.0)
+
+ /**
+ * Set the standard deviation of initial coefficients.
+ * Default is 0.01.
+ *
+ * @group setParam
+ */
+ @Since("3.0.0")
+ def setInitStd(value: Double): this.type = set(initStd, value)
+ setDefault(initStd -> 0.01)
+
+ /**
+ * Set the maximum number of iterations.
+ * Default is 100.
+ *
+ * @group
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r344544387
##
File path:
mllib/src/main/scala/org/apache/spark/ml/regression/FactorizationMachines.scala
##
@@ -0,0 +1,757 @@
+/*
+ * 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[regression] trait FactorizationMachinesParams
+ extends PredictorParams
+ with HasMaxIter with HasStepSize with HasTol with HasSolver with HasLoss {
+
+ import FactorizationMachines._
+
+ /**
+ * 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 parameter of l2-regularization term, " +
+ "which prevents overfitting by adding sum of squares of all the
parameters",
+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)
+
+ /**
+ * The solver algorithm for optimization.
+ * Supported
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r344541468
##
File path:
mllib/src/main/scala/org/apache/spark/ml/regression/FMRegressor.scala
##
@@ -0,0 +1,786 @@
+/*
+ * 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.regression.FactorizationMachines._
+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.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)
+
+ /**
+ * 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",
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r344541102
##
File path:
mllib/src/main/scala/org/apache/spark/ml/regression/FMRegressor.scala
##
@@ -0,0 +1,786 @@
+/*
+ * 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.regression.FactorizationMachines._
+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.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",
Review comment:
> why not extending `HasRegParam`?
Because I want to change `regParam` doc (FM regParams only for L2), but
`HasRegParam` has `final` tag, I can't override 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.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
-
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r344536192
##
File path:
mllib/src/main/scala/org/apache/spark/ml/classification/FMClassifier.scala
##
@@ -0,0 +1,326 @@
+/*
+ * 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.classification
+
+import org.apache.hadoop.fs.Path
+
+import org.apache.spark.annotation.Since
+import org.apache.spark.internal.Logging
+import org.apache.spark.ml.linalg._
+import org.apache.spark.ml.param._
+import org.apache.spark.ml.regression.{FactorizationMachines,
FactorizationMachinesParams}
+import org.apache.spark.ml.regression.FactorizationMachines._
+import org.apache.spark.ml.util._
+import org.apache.spark.ml.util.Instrumentation.instrumented
+import org.apache.spark.mllib.linalg.{Vector => OldVector}
+import org.apache.spark.mllib.linalg.VectorImplicits._
+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 FMClassifier.
+ */
+private[classification] trait FMClassifierParams extends
ProbabilisticClassifierParams
+ with FactorizationMachinesParams {
+}
+
+/**
+ * Factorization Machines learning algorithm for classification.
+ * It supports normal gradient descent and AdamW solver.
+ *
+ * The implementation is based upon:
+ * https://www.csie.ntu.edu.tw/~b97053/paper/Rendle2010FM.pdf;>
+ * S. Rendle. "Factorization machines" 2010.
+ *
+ * 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 classification model uses logistic loss which can be solved by gradient
descent method, and
+ * regularization terms like L2 are usually added to the loss function to
prevent overfitting.
+ *
+ * @note Multiclass labels are not currently supported.
+ */
+@Since("3.0.0")
+class FMClassifier @Since("3.0.0") (
+@Since("3.0.0") override val uid: String)
+ extends ProbabilisticClassifier[Vector, FMClassifier, FMClassifierModel]
+ with FactorizationMachines with FMClassifierParams with
DefaultParamsWritable with Logging {
+
+ @Since("3.0.0")
+ def this() = this(Identifiable.randomUID("fmc"))
+
+ /**
+ * 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 = set(miniBatchFraction,
value)
+ setDefault(miniBatchFraction -> 1.0)
+
+ /**
+ * Set the standard deviation of initial coefficients.
+ * Default is 0.01.
+ *
+ * @group setParam
+ */
+ @Since("3.0.0")
+ def setInitStd(value: Double): this.type = set(initStd, value)
+ setDefault(initStd -> 0.01)
+
+ /**
+ * Set the maximum number of iterations.
+ * Default is 100.
+ *
+ * @group
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r344534332
##
File path:
mllib/src/main/scala/org/apache/spark/ml/classification/FMClassifier.scala
##
@@ -0,0 +1,326 @@
+/*
+ * 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.classification
+
+import org.apache.hadoop.fs.Path
+
+import org.apache.spark.annotation.Since
+import org.apache.spark.internal.Logging
+import org.apache.spark.ml.linalg._
+import org.apache.spark.ml.param._
+import org.apache.spark.ml.regression.{FactorizationMachines,
FactorizationMachinesParams}
+import org.apache.spark.ml.regression.FactorizationMachines._
+import org.apache.spark.ml.util._
+import org.apache.spark.ml.util.Instrumentation.instrumented
+import org.apache.spark.mllib.linalg.{Vector => OldVector}
+import org.apache.spark.mllib.linalg.VectorImplicits._
+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 FMClassifier.
+ */
+private[classification] trait FMClassifierParams extends
ProbabilisticClassifierParams
+ with FactorizationMachinesParams {
+}
+
+/**
+ * Factorization Machines learning algorithm for classification.
+ * It supports normal gradient descent and AdamW solver.
+ *
+ * The implementation is based upon:
+ * https://www.csie.ntu.edu.tw/~b97053/paper/Rendle2010FM.pdf;>
+ * S. Rendle. "Factorization machines" 2010.
+ *
+ * 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 classification model uses logistic loss which can be solved by gradient
descent method, and
+ * regularization terms like L2 are usually added to the loss function to
prevent overfitting.
+ *
+ * @note Multiclass labels are not currently supported.
+ */
+@Since("3.0.0")
+class FMClassifier @Since("3.0.0") (
+@Since("3.0.0") override val uid: String)
+ extends ProbabilisticClassifier[Vector, FMClassifier, FMClassifierModel]
+ with FactorizationMachines with FMClassifierParams with
DefaultParamsWritable with Logging {
+
+ @Since("3.0.0")
+ def this() = this(Identifiable.randomUID("fmc"))
+
+ /**
+ * 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 = set(miniBatchFraction,
value)
+ setDefault(miniBatchFraction -> 1.0)
+
+ /**
+ * Set the standard deviation of initial coefficients.
+ * Default is 0.01.
+ *
+ * @group setParam
+ */
+ @Since("3.0.0")
+ def setInitStd(value: Double): this.type = set(initStd, value)
+ setDefault(initStd -> 0.01)
+
+ /**
+ * Set the maximum number of iterations.
+ * Default is 100.
+ *
+ * @group
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r342371780
##
File path:
mllib/src/test/scala/org/apache/spark/ml/classification/FMClassifierSuite.scala
##
@@ -0,0 +1,143 @@
+/*
+ * 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.classification
+
+import
org.apache.spark.ml.classification.LogisticRegressionSuite.generateLogisticInput
+import org.apache.spark.ml.linalg.{DenseVector, Vector, Vectors}
+import org.apache.spark.ml.param.ParamsSuite
+import org.apache.spark.ml.regression.FMRegressorSuite._
+import org.apache.spark.ml.util._
+import org.apache.spark.ml.util.TestingUtils._
+import org.apache.spark.sql.{DataFrame, Row}
+
+class FMClassifierSuite extends MLTest with DefaultReadWriteTest {
Review comment:
I want to compare with libFM, but randomly generate a binary dataset, FM
can't fit the dataset perfectly. Generally, I get very small loss value, but
the parameters are not exactly same every time (also in libFM).
My current test method is that for mse to generate a feature-interaction
dataset by bias/linearVector/factorMatrix, FMRegressor needs to fit this
dataset perfectly, bias/linearVector needs same with real bias/linearVector,
every `` needs same that calculates by factorMatrix.
However, for FMClassifier, it is difficult to generate a binary
feature-interaction dataset, so for the FMClassifier, I don't verify the
correctness of interaction logic (that is verify in mse), only verify the
correctness of logloss logic.
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:
us...@infra.apache.org
With regards,
Apache Git Services
-
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r342371780
##
File path:
mllib/src/test/scala/org/apache/spark/ml/classification/FMClassifierSuite.scala
##
@@ -0,0 +1,143 @@
+/*
+ * 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.classification
+
+import
org.apache.spark.ml.classification.LogisticRegressionSuite.generateLogisticInput
+import org.apache.spark.ml.linalg.{DenseVector, Vector, Vectors}
+import org.apache.spark.ml.param.ParamsSuite
+import org.apache.spark.ml.regression.FMRegressorSuite._
+import org.apache.spark.ml.util._
+import org.apache.spark.ml.util.TestingUtils._
+import org.apache.spark.sql.{DataFrame, Row}
+
+class FMClassifierSuite extends MLTest with DefaultReadWriteTest {
Review comment:
I want to compare with libFM, but randomly generate a binary dataset, FM
can't fit the dataset perfectly. Generally, I get very small loss value, but
the parameters are not exactly same every time (also in libFM).
My current test method is that for mse to generate a feature-interaction
dataset, FMRegressor needs to fit this dataset perfectly, bias/linearVector
needs same with real bias/linearVector, every `` needs same that
calculates by factorMatrix.
However, for FMClassifier, it is difficult to generate a binary
feature-interaction dataset, so for the FMClassifier, I don't verify the
correctness of interaction logic (that is verify in mse), only verify the
correctness of logloss logic.
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:
us...@infra.apache.org
With regards,
Apache Git Services
-
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r342371780
##
File path:
mllib/src/test/scala/org/apache/spark/ml/classification/FMClassifierSuite.scala
##
@@ -0,0 +1,143 @@
+/*
+ * 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.classification
+
+import
org.apache.spark.ml.classification.LogisticRegressionSuite.generateLogisticInput
+import org.apache.spark.ml.linalg.{DenseVector, Vector, Vectors}
+import org.apache.spark.ml.param.ParamsSuite
+import org.apache.spark.ml.regression.FMRegressorSuite._
+import org.apache.spark.ml.util._
+import org.apache.spark.ml.util.TestingUtils._
+import org.apache.spark.sql.{DataFrame, Row}
+
+class FMClassifierSuite extends MLTest with DefaultReadWriteTest {
Review comment:
I want to compare with libFM, but randomly generate a binary dataset, FM
can't fit the dataset perfectly. Generally, I get very small loss value, but
the parameters are not exactly same every time (also in libFM).
My current test method is that for mse to generate a feature-interaction
dataset, FMRegressor needs to fit this dataset perfectly, bias/linearVector
needs same, every `` needs same that calculates by factorMatrix.
However, for FMClassifier, it is difficult to generate a binary
feature-interaction dataset, so for the FMClassifier, I don't verify the
correctness of interaction logic (that is verify in mse), only verify the
correctness of logloss logic.
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:
us...@infra.apache.org
With regards,
Apache Git Services
-
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r341465703
##
File path:
mllib/src/main/scala/org/apache/spark/ml/classification/FMClassifier.scala
##
@@ -0,0 +1,360 @@
+/*
+ * 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.classification
+
+import org.apache.hadoop.fs.Path
+
+import org.apache.spark.annotation.Since
+import org.apache.spark.internal.Logging
+import org.apache.spark.ml.linalg._
+import org.apache.spark.ml.param._
+import org.apache.spark.ml.regression._
+import org.apache.spark.ml.util._
+import org.apache.spark.ml.util.Instrumentation.instrumented
+import org.apache.spark.mllib.linalg.{Vector => OldVector}
+import org.apache.spark.mllib.linalg.VectorImplicits._
+import org.apache.spark.mllib.optimization.GradientDescent
+import org.apache.spark.mllib.regression.{LabeledPoint => OldLabeledPoint}
+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 FMClassifier.
+ */
+private[classification] trait FMClassifierParams extends
ProbabilisticClassifierParams
+ with FactorizationMachinesParams {
+}
+
+/**
+ * Factorization Machines learning algorithm for classification.
+ * It supports normal gradient descent and AdamW solver.
+ *
+ * The implementation is based upon:
+ * https://www.csie.ntu.edu.tw/~b97053/paper/Rendle2010FM.pdf;>
+ * S. Rendle. "Factorization machines" 2010.
+ *
+ * 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 classification model uses logistic loss which can be solved by gradient
descent method, and
+ * regularization terms like L2 are usually added to the loss function to
prevent overfitting.
+ *
+ * @note Multiclass labels are not currently supported.
+ */
+@Since("3.0.0")
+class FMClassifier @Since("3.0.0") (
+ @Since("3.0.0") override val uid: String)
+ extends ProbabilisticClassifier[Vector, FMClassifier, FMClassifierModel]
+ with FMClassifierParams with DefaultParamsWritable with Logging {
+
+ import
org.apache.spark.ml.regression.BaseFactorizationMachinesGradient.{LogisticLoss,
parseLoss}
+ import org.apache.spark.ml.regression.FMRegressor.initCoefficients
+
+ @Since("3.0.0")
+ def this() = this(Identifiable.randomUID("fmc"))
+
+ /**
+ * 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,
+ s"Fraction for mini-batch SGD must be in range (0, 1] but got $value")
+set(miniBatchFraction, value)
+ }
+ setDefault(miniBatchFraction -> 1.0)
+
+ /**
+ * Set the standard
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r340405616
##
File path:
mllib/src/main/scala/org/apache/spark/ml/regression/FactorizationMachines.scala
##
@@ -0,0 +1,757 @@
+/*
+ * 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[regression] trait FactorizationMachinesParams
+ extends PredictorParams
+ with HasMaxIter with HasStepSize with HasTol with HasSolver with HasLoss {
+
+ import FactorizationMachines._
+
+ /**
+ * 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 parameter of l2-regularization term, " +
+ "which prevents overfitting by adding sum of squares of all the
parameters",
+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)
+
+ /**
+ * The solver algorithm for optimization.
+ * Supported
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r339962146
##
File path:
mllib/src/main/scala/org/apache/spark/ml/regression/FactorizationMachines.scala
##
@@ -0,0 +1,757 @@
+/*
+ * 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[regression] trait FactorizationMachinesParams
+ extends PredictorParams
+ with HasMaxIter with HasStepSize with HasTol with HasSolver with HasLoss {
+
+ import FactorizationMachines._
+
+ /**
+ * 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 parameter of l2-regularization term, " +
+ "which prevents overfitting by adding sum of squares of all the
parameters",
+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)
+
+ /**
+ * The solver algorithm for optimization.
+ * Supported
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r339938579
##
File path:
mllib/src/main/scala/org/apache/spark/ml/regression/FactorizationMachines.scala
##
@@ -0,0 +1,757 @@
+/*
+ * 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[regression] trait FactorizationMachinesParams
+ extends PredictorParams
+ with HasMaxIter with HasStepSize with HasTol with HasSolver with HasLoss {
+
+ import FactorizationMachines._
+
+ /**
+ * 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 parameter of l2-regularization term, " +
+ "which prevents overfitting by adding sum of squares of all the
parameters",
+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)
+
+ /**
+ * The solver algorithm for optimization.
+ * Supported
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r339927384
##
File path:
mllib/src/main/scala/org/apache/spark/ml/regression/FactorizationMachines.scala
##
@@ -0,0 +1,757 @@
+/*
+ * 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[regression] trait FactorizationMachinesParams
+ extends PredictorParams
+ with HasMaxIter with HasStepSize with HasTol with HasSolver with HasLoss {
+
+ import FactorizationMachines._
+
+ /**
+ * 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 parameter of l2-regularization term, " +
+ "which prevents overfitting by adding sum of squares of all the
parameters",
+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)
+
+ /**
+ * The solver algorithm for optimization.
+ * Supported
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r339925551
##
File path:
mllib/src/main/scala/org/apache/spark/ml/regression/FactorizationMachines.scala
##
@@ -0,0 +1,757 @@
+/*
+ * 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[regression] trait FactorizationMachinesParams
+ extends PredictorParams
+ with HasMaxIter with HasStepSize with HasTol with HasSolver with HasLoss {
+
+ import FactorizationMachines._
+
+ /**
+ * 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 parameter of l2-regularization term, " +
+ "which prevents overfitting by adding sum of squares of all the
parameters",
+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)
+
+ /**
+ * The solver algorithm for optimization.
+ * Supported
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r339925551
##
File path:
mllib/src/main/scala/org/apache/spark/ml/regression/FactorizationMachines.scala
##
@@ -0,0 +1,757 @@
+/*
+ * 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[regression] trait FactorizationMachinesParams
+ extends PredictorParams
+ with HasMaxIter with HasStepSize with HasTol with HasSolver with HasLoss {
+
+ import FactorizationMachines._
+
+ /**
+ * 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 parameter of l2-regularization term, " +
+ "which prevents overfitting by adding sum of squares of all the
parameters",
+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)
+
+ /**
+ * The solver algorithm for optimization.
+ * Supported
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r339923757
##
File path:
mllib/src/main/scala/org/apache/spark/ml/regression/FactorizationMachines.scala
##
@@ -0,0 +1,757 @@
+/*
+ * 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[regression] trait FactorizationMachinesParams
+ extends PredictorParams
+ with HasMaxIter with HasStepSize with HasTol with HasSolver with HasLoss {
+
+ import FactorizationMachines._
+
+ /**
+ * 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 parameter of l2-regularization term, " +
+ "which prevents overfitting by adding sum of squares of all the
parameters",
+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)
+
+ /**
+ * The solver algorithm for optimization.
+ * Supported
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r338951244
##
File path: python/pyspark/ml/regression.py
##
@@ -2126,6 +2127,181 @@ def __repr__(self):
return self._call_java("toString")
+@inherit_doc
+class FactorizationMachines(JavaPredictor, HasMaxIter, HasStepSize, HasTol,
HasSolver, HasLoss,
+JavaMLWritable, JavaMLReadable):
+"""
+Factorization Machines.
+
+loss Supports:
+
+* logisticLoss (default)
Review comment:
> Dumb question but does this render correctly in python docs? I'm not sure
this is how bulleted lists work, but am not sure
I refer to the LinearRegression docs (regression.py: line51). In my
experience, python \_\_doc\_\_ just a string, it won't render as a list. But I
have not idea, whether spark will use \_\_doc\_\_ to generate docs html.
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:
us...@infra.apache.org
With regards,
Apache Git Services
-
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r338951244
##
File path: python/pyspark/ml/regression.py
##
@@ -2126,6 +2127,181 @@ def __repr__(self):
return self._call_java("toString")
+@inherit_doc
+class FactorizationMachines(JavaPredictor, HasMaxIter, HasStepSize, HasTol,
HasSolver, HasLoss,
+JavaMLWritable, JavaMLReadable):
+"""
+Factorization Machines.
+
+loss Supports:
+
+* logisticLoss (default)
Review comment:
> Dumb question but does this render correctly in python docs? I'm not sure
this is how bulleted lists work, but am not sure
I refer to the LinearRegression docs (regression.py: line51). In my
experience, python doc just a string, it won't render as a list. But I
have not idea, whether spark will use doc to generate docs html.
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:
us...@infra.apache.org
With regards,
Apache Git Services
-
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r338951244
##
File path: python/pyspark/ml/regression.py
##
@@ -2126,6 +2127,181 @@ def __repr__(self):
return self._call_java("toString")
+@inherit_doc
+class FactorizationMachines(JavaPredictor, HasMaxIter, HasStepSize, HasTol,
HasSolver, HasLoss,
+JavaMLWritable, JavaMLReadable):
+"""
+Factorization Machines.
+
+loss Supports:
+
+* logisticLoss (default)
Review comment:
> Dumb question but does this render correctly in python docs? I'm not sure
this is how bulleted lists work, but am not sure
I refer to the LinearRegression docs (regression.py: line51). In my
experience, python "__doc__" just a string, it won't render as a list. But I
have not idea, whether spark will use "__doc__" to generate docs html.
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:
us...@infra.apache.org
With regards,
Apache Git Services
-
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r338951244
##
File path: python/pyspark/ml/regression.py
##
@@ -2126,6 +2127,181 @@ def __repr__(self):
return self._call_java("toString")
+@inherit_doc
+class FactorizationMachines(JavaPredictor, HasMaxIter, HasStepSize, HasTol,
HasSolver, HasLoss,
+JavaMLWritable, JavaMLReadable):
+"""
+Factorization Machines.
+
+loss Supports:
+
+* logisticLoss (default)
Review comment:
> Dumb question but does this render correctly in python docs? I'm not sure
this is how bulleted lists work, but am not sure
I refer to the LinearRegression docs (regression.py: line51). In my
experience, python __doc__ just a string, it won't render as a list. But I have
not idea, whether spark will use __doc__ to generate docs html.
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:
us...@infra.apache.org
With regards,
Apache Git Services
-
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r338864620
##
File path:
mllib/src/main/scala/org/apache/spark/ml/regression/FactorizationMachines.scala
##
@@ -0,0 +1,839 @@
+/*
+ * 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, 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
Review comment:
FM paper: S. Rendle, “Factorization machines,” in Proceedings of IEEE
International Conference on Data Mining (ICDM), pp. 995–1000, 2010.
https://www.csie.ntu.edu.tw/~b97053/paper/Rendle2010FM.pdf
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:
us...@infra.apache.org
With regards,
Apache Git Services
-
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r338865533
##
File path:
mllib/src/main/scala/org/apache/spark/ml/regression/FactorizationMachines.scala
##
@@ -0,0 +1,839 @@
+/*
+ * 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, 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
Review comment:
> Is there any paper you can link to to explain the implementation? or, just
a few paragraphs about what the implementation does?
FM paper: S. Rendle, “Factorization machines,” in Proceedings of IEEE
International Conference on Data Mining (ICDM), pp. 995–1000, 2010.
https://www.csie.ntu.edu.tw/~b97053/paper/Rendle2010FM.pdf
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:
us...@infra.apache.org
With regards,
Apache Git Services
-
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r338865488
##
File path:
mllib/src/main/scala/org/apache/spark/ml/regression/FactorizationMachines.scala
##
@@ -0,0 +1,839 @@
+/*
+ * 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, 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[regression] trait FactorizationMachinesParams
+ extends PredictorParams
+ with HasMaxIter with HasStepSize with HasTol with HasSolver with HasLoss {
+
+ import FactorizationMachines._
+
+ /**
+ * Param for dimensionality of the factors (= 0)
+ * @group param
+ */
+ @Since("3.0.0")
+ final val numFactors: IntParam = new IntParam(this, "numFactors",
+"dimensionality of the factorization")
+
+ /** @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",
+"regularization for L2")
+
+ /** @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",
+"mini-batch fraction")
+
+ /** @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")
+
+ /** @group getParam */
+ @Since("3.0.0")
+ final def getInitStd: Double = $(initStd)
+
+ /**
+ * The solver algorithm for optimization.
+ * Supported options: "gd", "adamW".
+ * Default: "adamW"
+ *
+ * @group param
+ */
+ @Since("3.0.0")
+ final override val solver: Param[String] = new Param[String](this, "solver",
+"The solver algorithm for optimization. Supported options: " +
+ s"${supportedSolvers.mkString(", ")}. (Default adamW)",
+ParamValidators.inArray[String](supportedSolvers))
+
+ /**
+ * The loss function to
[GitHub] [spark] mob-ai commented on a change in pull request #26124: [SPARK-29224][ML]Implement Factorization Machines as a ml-pipeline component
mob-ai 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_r338864620
##
File path:
mllib/src/main/scala/org/apache/spark/ml/regression/FactorizationMachines.scala
##
@@ -0,0 +1,839 @@
+/*
+ * 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, 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
Review comment:
FM paper: S. Rendle, “Factorization machines,” in Proceedings of IEEE
International Conference on Data Mining (ICDM), pp. 995–1000, 2010.
https://www.csie.ntu.edu.tw/~b97053/paper/Rendle2010FM.pdf
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:
us...@infra.apache.org
With regards,
Apache Git Services
-
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
