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 Mon, 04 May 2015 14:34:06 GMT

    [ https://issues.apache.org/jira/browse/FLINK-1807?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14526682#comment-14526682
] 

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

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

    https://github.com/apache/flink/pull/613#discussion_r29588896
  
    --- Diff: flink-staging/flink-ml/src/main/scala/org/apache/flink/ml/optimization/LossFunction.scala
---
    @@ -0,0 +1,101 @@
    +/*
    + * 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.ml.common.{WeightVector, LabeledVector}
    +import org.apache.flink.ml.math.{Vector => FlinkVector, BLAS}
    +
    +
    +abstract class LossFunction extends Serializable{
    +
    +
    +  /** Calculates the loss for a given prediction/truth pair
    +    *
    +    * @param prediction The predicted value
    +    * @param truth The true value
    +    */
    +  protected def loss(prediction: Double, truth: Double): Double
    +
    +  /** Calculates the derivative of the loss function with respect to the prediction
    +    *
    +    * @param prediction The predicted value
    +    * @param truth The true value
    +    */
    +  protected def lossDerivative(prediction: Double, truth: Double): Double
    +
    +  /** Compute the gradient and the loss for the given data.
    +    * The provided cumGradient is updated in place.
    +    *
    +    * @param example The features and the label associated with the example
    +    * @param weights The current weight vector
    +    * @param cumGradient The vector to which the gradient will be added to, in place.
    +    * @return A tuple containing the computed loss as its first element and a the loss
derivative as
    +    *         its second element.
    +    */
    +  def lossAndGradient(
    +      example: LabeledVector,
    +      weights: WeightVector,
    +      cumGradient: FlinkVector,
    +      regType:  RegularizationType,
    +      regParameter: Double):  (Double, Double) = {
    +    val features = example.vector
    +    val label = example.label
    +    // TODO(tvas): We could also provide for the case where we don't want an intercept
value
    +    // i.e. data already centered
    +    val prediction = BLAS.dot(features, weights.weights) + weights.intercept
    --- End diff --
    
    Forgot about this, will fix and push again.


> 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