diff options
author | Xiangrui Meng <meng@databricks.com> | 2015-08-12 23:04:59 -0700 |
---|---|---|
committer | Xiangrui Meng <meng@databricks.com> | 2015-08-12 23:04:59 -0700 |
commit | 68f99571492f67596b3656e9f076deeb96616f4a (patch) | |
tree | be2248de6c66c50ce13561c67c18af7e3ae86f2e /mllib/src/main/scala | |
parent | d0b18919d16e6a2f19159516bd2767b60b595279 (diff) | |
download | spark-68f99571492f67596b3656e9f076deeb96616f4a.tar.gz spark-68f99571492f67596b3656e9f076deeb96616f4a.tar.bz2 spark-68f99571492f67596b3656e9f076deeb96616f4a.zip |
[SPARK-9918] [MLLIB] remove runs from k-means and rename epsilon to tol
This requires some discussion. I'm not sure whether `runs` is a useful parameter. It certainly complicates the implementation. We might want to optimize the k-means implementation with block matrix operations. In this case, having `runs` may not be worth the trade-off. Also it increases the communication cost in a single job, which might cause other issues.
This PR also renames `epsilon` to `tol` to have consistent naming among algorithms. The Python constructor is updated to include all parameters.
jkbradley yu-iskw
Author: Xiangrui Meng <meng@databricks.com>
Closes #8148 from mengxr/SPARK-9918 and squashes the following commits:
149b9e5 [Xiangrui Meng] fix constructor in Python and rename epsilon to tol
3cc15b3 [Xiangrui Meng] fix test and change initStep to initSteps in python
a0a0274 [Xiangrui Meng] remove runs from k-means in the pipeline API
Diffstat (limited to 'mllib/src/main/scala')
-rw-r--r-- | mllib/src/main/scala/org/apache/spark/ml/clustering/KMeans.scala | 51 |
1 files changed, 10 insertions, 41 deletions
diff --git a/mllib/src/main/scala/org/apache/spark/ml/clustering/KMeans.scala b/mllib/src/main/scala/org/apache/spark/ml/clustering/KMeans.scala index dc192add6c..47a18cdb31 100644 --- a/mllib/src/main/scala/org/apache/spark/ml/clustering/KMeans.scala +++ b/mllib/src/main/scala/org/apache/spark/ml/clustering/KMeans.scala @@ -18,8 +18,8 @@ package org.apache.spark.ml.clustering import org.apache.spark.annotation.Experimental -import org.apache.spark.ml.param.{Param, Params, IntParam, DoubleParam, ParamMap} -import org.apache.spark.ml.param.shared.{HasFeaturesCol, HasMaxIter, HasPredictionCol, HasSeed} +import org.apache.spark.ml.param.{Param, Params, IntParam, ParamMap} +import org.apache.spark.ml.param.shared._ import org.apache.spark.ml.util.{Identifiable, SchemaUtils} import org.apache.spark.ml.{Estimator, Model} import org.apache.spark.mllib.clustering.{KMeans => MLlibKMeans, KMeansModel => MLlibKMeansModel} @@ -27,14 +27,13 @@ import org.apache.spark.mllib.linalg.{Vector, VectorUDT} import org.apache.spark.sql.functions.{col, udf} import org.apache.spark.sql.types.{IntegerType, StructType} import org.apache.spark.sql.{DataFrame, Row} -import org.apache.spark.util.Utils /** * Common params for KMeans and KMeansModel */ -private[clustering] trait KMeansParams - extends Params with HasMaxIter with HasFeaturesCol with HasSeed with HasPredictionCol { +private[clustering] trait KMeansParams extends Params with HasMaxIter with HasFeaturesCol + with HasSeed with HasPredictionCol with HasTol { /** * Set the number of clusters to create (k). Must be > 1. Default: 2. @@ -46,31 +45,6 @@ private[clustering] trait KMeansParams def getK: Int = $(k) /** - * Param the number of runs of the algorithm to execute in parallel. We initialize the algorithm - * this many times with random starting conditions (configured by the initialization mode), then - * return the best clustering found over any run. Must be >= 1. Default: 1. - * @group param - */ - final val runs = new IntParam(this, "runs", - "number of runs of the algorithm to execute in parallel", (value: Int) => value >= 1) - - /** @group getParam */ - def getRuns: Int = $(runs) - - /** - * Param the distance threshold within which we've consider centers to have converged. - * If all centers move less than this Euclidean distance, we stop iterating one run. - * Must be >= 0.0. Default: 1e-4 - * @group param - */ - final val epsilon = new DoubleParam(this, "epsilon", - "distance threshold within which we've consider centers to have converge", - (value: Double) => value >= 0.0) - - /** @group getParam */ - def getEpsilon: Double = $(epsilon) - - /** * Param for the initialization algorithm. This can be either "random" to choose random points as * initial cluster centers, or "k-means||" to use a parallel variant of k-means++ * (Bahmani et al., Scalable K-Means++, VLDB 2012). Default: k-means||. @@ -136,9 +110,9 @@ class KMeansModel private[ml] ( /** * :: Experimental :: - * K-means clustering with support for multiple parallel runs and a k-means++ like initialization - * mode (the k-means|| algorithm by Bahmani et al). When multiple concurrent runs are requested, - * they are executed together with joint passes over the data for efficiency. + * K-means clustering with support for k-means|| initialization proposed by Bahmani et al. + * + * @see [[http://dx.doi.org/10.14778/2180912.2180915 Bahmani et al., Scalable k-means++.]] */ @Experimental class KMeans(override val uid: String) extends Estimator[KMeansModel] with KMeansParams { @@ -146,10 +120,9 @@ class KMeans(override val uid: String) extends Estimator[KMeansModel] with KMean setDefault( k -> 2, maxIter -> 20, - runs -> 1, initMode -> MLlibKMeans.K_MEANS_PARALLEL, initSteps -> 5, - epsilon -> 1e-4) + tol -> 1e-4) override def copy(extra: ParamMap): KMeans = defaultCopy(extra) @@ -174,10 +147,7 @@ class KMeans(override val uid: String) extends Estimator[KMeansModel] with KMean def setMaxIter(value: Int): this.type = set(maxIter, value) /** @group setParam */ - def setRuns(value: Int): this.type = set(runs, value) - - /** @group setParam */ - def setEpsilon(value: Double): this.type = set(epsilon, value) + def setTol(value: Double): this.type = set(tol, value) /** @group setParam */ def setSeed(value: Long): this.type = set(seed, value) @@ -191,8 +161,7 @@ class KMeans(override val uid: String) extends Estimator[KMeansModel] with KMean .setInitializationSteps($(initSteps)) .setMaxIterations($(maxIter)) .setSeed($(seed)) - .setEpsilon($(epsilon)) - .setRuns($(runs)) + .setEpsilon($(tol)) val parentModel = algo.run(rdd) val model = new KMeansModel(uid, parentModel) copyValues(model) |