spark-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From sethah <...@git.apache.org>
Subject [GitHub] spark pull request: [SPARK-9835] [ML] IterativelyReweightedLeastSq...
Date Fri, 08 Jan 2016 00:13:23 GMT
Github user sethah commented on a diff in the pull request:

    https://github.com/apache/spark/pull/10639#discussion_r49144625
  
    --- Diff: mllib/src/main/scala/org/apache/spark/ml/optim/GLMFamilies.scala ---
    @@ -0,0 +1,123 @@
    +/*
    + * 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.optim
    +
    +import org.apache.spark.rdd.RDD
    +
    +/**
    + * A description of the error distribution and link function to be used in the model.
    + * @param link a link function instance
    + */
    +private[ml] abstract class Family(val link: Link) extends Serializable {
    +
    +  /**
    +   * Starting value for mu in the IRLS algorithm.
    +   */
    +  def startingMu(y: Double, yMean: Double): Double = (y + yMean) / 2.0
    +
    +  /**
    +   * Deviance of (y, mu) pair.
    +   * Deviance is usually defined as twice the loglikelihood ratio.
    +   */
    +  def deviance(y: RDD[Double], mu: RDD[Double]): Double
    +
    +  /** Weights for IRLS steps. */
    +  def weights(mu: Double): Double
    +
    +  /** The working dependent variable. */
    +  def z(y: Double, mu: Double, eta: Double): Double
    +}
    +
    +/**
    + * Binomial exponential family distribution.
    + * The default link for the Binomial family is the logit link.
    + * @param link a link function instance
    + */
    +private[ml] class Binomial(link: Link = new Logit) extends Family(link) {
    +
    +  override def startingMu(y: Double, yMean: Double): Double = (y + 0.5) / 2.0
    +
    +  override def deviance(y: RDD[Double], mu: RDD[Double]): Double = {
    +    mu.zip(y).map { case (mu, y) =>
    +      val my = 1.0 - y
    +      y * math.log(math.max(y, 1.0) / mu) +
    +        my * math.log(math.max(my, 1.0) / (1.0 - mu))
    +    }.sum() * 2
    +  }
    +
    +  override def weights(mu: Double): Double = {
    +    mu * (1 - mu)
    +  }
    +
    +  override def z(y: Double, mu: Double, eta: Double): Double = {
    +    eta + (y - mu) * link.deriv(mu)
    +  }
    +}
    +
    +/**
    + * Poisson exponential family.
    + * The default link for the Poisson family is the log link.
    + * @param link a link function instance
    + */
    +private[ml] class Poisson(link: Link = new Logit) extends Family(link) {
    --- End diff --
    
    I believe the link function here should default to `Log` not `Logit`


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---

---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org


Mime
View raw message