flink-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF GitHub Bot (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (FLINK-1807) Stochastic gradient descent optimizer for ML library
Date Thu, 07 May 2015 09:40:59 GMT

    [ https://issues.apache.org/jira/browse/FLINK-1807?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14532341#comment-14532341
] 

ASF GitHub Bot commented on FLINK-1807:
---------------------------------------

Github user tillrohrmann commented on a diff in the pull request:

    https://github.com/apache/flink/pull/613#discussion_r29837516
  
    --- Diff: flink-staging/flink-ml/src/main/scala/org/apache/flink/ml/optimization/Solver.scala
---
    @@ -0,0 +1,146 @@
    +/*
    + * Licensed to the Apache Software Foundation (ASF) under one
    + * or more contributor license agreements.  See the NOTICE file
    + * distributed with this work for additional information
    + * regarding copyright ownership.  The ASF licenses this file
    + * to you under the Apache License, Version 2.0 (the
    + * "License"); you may not use this file except in compliance
    + * with the License.  You may obtain a copy of the License at
    + *
    + *     http://www.apache.org/licenses/LICENSE-2.0
    + *
    + * Unless required by applicable law or agreed to in writing, software
    + * distributed under the License is distributed on an "AS IS" BASIS,
    + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
    + * See the License for the specific language governing permissions and
    + * limitations under the License.
    + */
    +
    +package org.apache.flink.ml.optimization
    +
    +import org.apache.flink.api.scala.DataSet
    +import org.apache.flink.ml.common._
    +import org.apache.flink.ml.math.{Vector => FlinkVector, BLAS, DenseVector}
    +import org.apache.flink.api.scala._
    +import org.apache.flink.ml.optimization.IterativeSolver._
    +import org.apache.flink.ml.optimization.Solver._
    +
    +/** Base class for optimization algorithms
    + *
    + */
    +abstract class Solver extends Serializable with WithParameters {
    +
    +  /** Provides a solution for the given optimization problem
    +    *
    +    * @param data A Dataset of LabeledVector (input, output) pairs
    +    * @param initialWeights The initial weight that will be optimized
    +    * @return A Vector of weights optimized to the given problem
    +    */
    +  def optimize(data: DataSet[LabeledVector], initialWeights: Option[DataSet[WeightVector]]):
    +  DataSet[WeightVector]
    +  // TODO(tvas): Maybe we want to pass a WeightVector directly here, instead of a
    +  // DataSet[WeightVector]
    +
    +  /** Creates a DataSet with one zero vector. The zero vector has dimension d, which
is given
    +    * by the dimensionDS.
    +    *
    +    * @param dimensionDS DataSet with one element d, denoting the dimension of the returned
zero
    +    *                    vector
    +    * @return DataSet of a zero vector of dimension d
    +    */
    +  def createInitialWeightVector(dimensionDS: DataSet[Int]):  DataSet[WeightVector] =
{
    +    dimensionDS.map {
    +      dimension =>
    +        val values = Array.fill(dimension)(0.0)
    +        new WeightVector(DenseVector(values), 0.0)
    +    }
    +  }
    +
    +  //Setters for parameters
    +  def setLossFunction(lossFunction: LossFunction): Solver = {
    +    parameters.add(LossFunction, lossFunction)
    +    this
    +  }
    +
    +  def setRegularizationType(regularization: RegularizationType): Solver = {
    +    parameters.add(RegularizationType, regularization)
    +    this
    +  }
    +
    +  def setRegularizationParameter(regularizationParameter: Double): Solver = {
    +    parameters.add(RegularizationParameter, regularizationParameter)
    +    this
    +  }
    +
    +  def setPredictionFunction(predictionFunction: PredictionFunction): Solver = {
    +    parameters.add(PredictionFunctionParam, predictionFunction)
    +    this
    +  }
    +}
    +
    +object Solver {
    +  // TODO(tvas): Does this belong in IterativeSolver instead?
    +  val WEIGHTVECTOR_BROADCAST = "weights_broadcast"
    +
    +  // Define parameters for Solver
    +  case object LossFunction extends Parameter[LossFunction] {
    +    // TODO(tvas): Should depend on problem, here is where differentiating between classification
    +    // and regression could become useful
    +    val defaultValue = Some(new SquaredLoss)
    +  }
    +
    +  case object RegularizationType extends Parameter[RegularizationType] {
    +    val defaultValue = Some(new NoRegularization)
    +  }
    +
    +  case object RegularizationParameter extends Parameter[Double] {
    +    val defaultValue = Some(0.0) // TODO(tvas): Properly initialize this, ensure Parameter
> 0!
    +  }
    +
    +  case object PredictionFunctionParam extends Parameter[PredictionFunction] {
    --- End diff --
    
    Consistent naming Param => Parameter


> Stochastic gradient descent optimizer for ML library
> ----------------------------------------------------
>
>                 Key: FLINK-1807
>                 URL: https://issues.apache.org/jira/browse/FLINK-1807
>             Project: Flink
>          Issue Type: Improvement
>          Components: Machine Learning Library
>            Reporter: Till Rohrmann
>            Assignee: Theodore Vasiloudis
>              Labels: ML
>
> Stochastic gradient descent (SGD) is a widely used optimization technique in different
ML algorithms. Thus, it would be helpful to provide a generalized SGD implementation which
can be instantiated with the respective gradient computation. Such a building block would
make the development of future algorithms easier.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message