forked from apache/spark
-
Notifications
You must be signed in to change notification settings - Fork 14
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Adding Power Iteration Clustering and Suite test
- Loading branch information
Showing
2 changed files
with
512 additions
and
4 deletions.
There are no files selected for viewing
248 changes: 244 additions & 4 deletions
248
mllib/src/main/scala/org/apache/spark/mllib/clustering/PIClustering.scala
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -1,8 +1,248 @@ | ||
/* | ||
* 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.mllib.clustering | ||
|
||
import org.apache.spark.graphx | ||
import org.apache.spark.SparkContext | ||
import org.apache.spark.rdd.RDD | ||
/** | ||
* Created by fan on 1/22/15. | ||
* Power Iteration Clustering | ||
* | ||
*/ | ||
class PIClustering { | ||
|
||
} | ||
object PIClustering { | ||
type DVector = Array[Double] | ||
type DEdge = Edge[Double] | ||
type LabeledPoint = (VertexId, DVector) | ||
type Points = Seq[LabeledPoint] | ||
type DGraph = Graph[Double, Double] | ||
type IndexedVector = (Long, DVector) | ||
val DefaultMinNormChange: Double = 1e-11 | ||
val DefaultSigma = 1.0 | ||
val DefaultIterations: Int = 20 | ||
val DefaultMinAffinity = 1e-11 | ||
val LA = SpectralClusteringUsingRdd.Linalg | ||
def cluster(sc: SparkContext, | ||
points: Points, | ||
nClusters: Int, | ||
nIterations: Int = DefaultIterations, | ||
sigma: Double = DefaultSigma, | ||
minAffinity: Double = DefaultMinAffinity) = { | ||
val nVertices = points.length | ||
val (wRdd, rowSums) = createNormalizedAffinityMatrix(sc, points, sigma) | ||
val initialVt = createInitialVector(sc, points.map(_._1), rowSums) | ||
val edgesRdd = createSparseEdgesRdd(sc, wRdd, minAffinity) | ||
val G = createGraphFromEdges(sc, edgesRdd, points.size, Some(initialVt)) | ||
getPrincipalEigen(sc, G) | ||
} | ||
/* | ||
vnorm[0]=2.019968019268192 | ||
Updating vertex[0] from 0.2592592592592593 to 0.2597973189724011 | ||
Updating vertex[1] from 0.19753086419753088 to 0.1695805301675885 | ||
Updating vertex[3] from 0.2654320987654321 to 0.27258531045499795 | ||
Updating vertex[2] from 0.2777777777777778 to 0.29803684040501227 | ||
*/ | ||
def createInitialVector(sc: SparkContext, | ||
labels: Seq[VertexId], | ||
rowSums: Seq[Double]) = { | ||
val volume = rowSums.fold(0.0) { | ||
_ + _ | ||
} | ||
val initialVt = labels.zip(rowSums.map(_ / volume)) | ||
initialVt | ||
} | ||
def createGraphFromEdges(sc: SparkContext, | ||
edgesRdd: RDD[DEdge], | ||
nPoints: Int, | ||
optInitialVt: Option[Seq[(VertexId, Double)]] = None) = { | ||
assert(nPoints > 0, "Must provide number of points from the original dataset") | ||
val G = if (optInitialVt.isDefined) { | ||
val initialVt = optInitialVt.get | ||
val vertsRdd = sc.parallelize(initialVt) | ||
Graph(vertsRdd, edgesRdd) | ||
} else { | ||
Graph.fromEdges(edgesRdd, -1.0) | ||
} | ||
G | ||
} | ||
val printMatrices = true | ||
def getPrincipalEigen(sc: SparkContext, | ||
G: DGraph, | ||
nIterations: Int = DefaultIterations, | ||
optMinNormChange: Option[Double] = None | ||
): (DGraph, Double, DVector) = { | ||
var priorNorm = Double.MaxValue | ||
var norm = Double.MaxValue | ||
var priorNormVelocity = Double.MaxValue | ||
var normVelocity = Double.MaxValue | ||
var normAccel = Double.MaxValue | ||
val DummyVertexId = -99L | ||
var vnorm: Double = -1.0 | ||
var outG: DGraph = null | ||
var prevG: DGraph = G | ||
val epsilon = optMinNormChange | ||
.getOrElse(1e-5 / G.vertices.count()) | ||
for (iter <- 0 until nIterations | ||
if Math.abs(normAccel) > epsilon) { | ||
val tmpEigen = prevG.aggregateMessages[Double](ctx => { | ||
ctx.sendToSrc(ctx.attr * ctx.srcAttr); | ||
ctx.sendToDst(ctx.attr * ctx.dstAttr) | ||
}, | ||
_ + _) | ||
println(s"tmpEigen[$iter]: ${tmpEigen.collect.mkString(",")}\n") | ||
val vnorm = | ||
prevG.vertices.map{ _._2}.fold(0.0) { case (sum, dval) => | ||
sum + Math.abs(dval) | ||
} | ||
println(s"vnorm[$iter]=$vnorm") | ||
outG = prevG.outerJoinVertices(tmpEigen) { case (vid, wval, optTmpEigJ) => | ||
val normedEig = optTmpEigJ.getOrElse { | ||
println("We got null estimated eigenvector element"); | ||
-1.0 | ||
} / vnorm | ||
println(s"Updating vertex[$vid] from $wval to $normedEig") | ||
normedEig | ||
} | ||
prevG = outG | ||
if (printMatrices) { | ||
val localVertices = outG.vertices.collect | ||
val graphSize = localVertices.size | ||
print(s"Vertices[$iter]: ${localVertices.mkString(",")}\n") | ||
} | ||
normVelocity = vnorm - priorNorm | ||
normAccel = normVelocity - priorNormVelocity | ||
println(s"normAccel[$iter]= $normAccel") | ||
priorNorm = vnorm | ||
priorNormVelocity = vnorm - priorNorm | ||
} | ||
(outG, vnorm, outG.vertices.collect.map { | ||
_._2 | ||
}) | ||
} | ||
// def printGraph(G: DGraph) = { | ||
// val collectedVerts = G.vertices.collect | ||
// val nVertices = collectedVerts.length | ||
// val msg = s"Graph Vertices:\n${printMatrix(collectedVerts, nVertices, nVertices)}" | ||
// } | ||
// | ||
def scalarDot(d1: DVector, d2: DVector) = { | ||
Math.sqrt(d1.zip(d2).foldLeft(0.0) { case (sum, (d1v, d2v)) => | ||
sum + d1v * d2v | ||
}) | ||
} | ||
def vectorDot(d1: DVector, d2: DVector) = { | ||
d1.zip(d2).map { case (d1v, d2v) => | ||
d1v * d2v | ||
} | ||
} | ||
def normVect(d1: DVector, d2: DVector) = { | ||
val scaldot = scalarDot(d1, d2) | ||
vectorDot(d1, d2).map { | ||
_ / scaldot | ||
} | ||
} | ||
def readVerticesfromFile(verticesFile: String): Points = { | ||
import scala.io.Source | ||
val vertices = Source.fromFile(verticesFile).getLines.map { l => | ||
val toks = l.split("\t") | ||
val arr = toks.slice(1, toks.length).map(_.toDouble) | ||
(toks(0).toLong, arr) | ||
}.toSeq | ||
println(s"Read in ${vertices.length} from $verticesFile") | ||
// println(vertices.map { case (x, arr) => s"($x,${arr.mkString(",")})"} | ||
// .mkString("[", ",\n", "]")) | ||
vertices | ||
} | ||
def gaussianDist(c1arr: DVector, c2arr: DVector, sigma: Double) = { | ||
val c1c2 = c1arr.zip(c2arr) | ||
val dist = Math.exp((0.5 / Math.pow(sigma, 2.0)) * c1c2.foldLeft(0.0) { | ||
case (dist: Double, (c1: Double, c2: Double)) => | ||
dist + Math.pow(c1 - c2, 2) | ||
}) | ||
dist | ||
} | ||
def createSparseEdgesRdd(sc: SparkContext, wRdd: RDD[IndexedVector], | ||
minAffinity: Double = DefaultMinAffinity) = { | ||
val labels = wRdd.map { case (vid, vect) => vid}.collect | ||
val edgesRdd = wRdd.flatMap { case (vid, vect) => | ||
for ((dval, ix) <- vect.zipWithIndex | ||
if Math.abs(dval) >= minAffinity) | ||
yield Edge(vid, labels(ix), dval) | ||
} | ||
edgesRdd | ||
} | ||
def createNormalizedAffinityMatrix(sc: SparkContext, points: Points, sigma: Double) = { | ||
val nVertices = points.length | ||
val rowSums = for (bcx <- 0 until nVertices) | ||
yield sc.accumulator[Double](bcx, s"ColCounts$bcx") | ||
val affinityRddNotNorm = sc.parallelize({ | ||
val ivect = new Array[IndexedVector](nVertices) | ||
var rsum = 0.0 | ||
for (i <- 0 until points.size) { | ||
ivect(i) = new IndexedVector(points(i)._1, new DVector(nVertices)) | ||
for (j <- 0 until points.size) { | ||
val dist = if (i != j) { | ||
gaussianDist(points(i)._2, points(j)._2, sigma) | ||
} else { | ||
0.0 | ||
} | ||
ivect(i)._2(j) = dist | ||
rsum += dist | ||
} | ||
rowSums(i) += rsum | ||
} | ||
ivect.zipWithIndex.map { case (vect, ix) => | ||
(ix, vect) | ||
} | ||
}, nVertices) | ||
val affinityRdd = affinityRddNotNorm.map { case (rowx, (vid, vect)) => | ||
(vid, vect.map { | ||
_ / rowSums(rowx).value | ||
}) | ||
} | ||
(affinityRdd, rowSums.map { | ||
_.value | ||
}) | ||
} | ||
def norm(vect: DVector): Double = { | ||
Math.sqrt(vect.foldLeft(0.0) { case (sum, dval) => sum + Math.pow(dval, 2)}) | ||
} | ||
def printMatrix(darr: Array[DVector], numRows: Int, numCols: Int): String = { | ||
val flattenedArr = darr.zipWithIndex.foldLeft(new DVector(numRows * numCols)) { | ||
case (flatarr, (row, indx)) => | ||
System.arraycopy(row, 0, flatarr, indx * numCols, numCols) | ||
flatarr | ||
} | ||
printMatrix(flattenedArr, numRows, numCols) | ||
} | ||
def printMatrix(darr: DVector, numRows: Int, numCols: Int): String = { | ||
val stride = (darr.length / numCols) | ||
val sb = new StringBuilder | ||
def leftJust(s: String, len: Int) = { | ||
" ".substring(0, len - Math.min(len, s.length)) + s | ||
} | ||
for (r <- 0 until numRows) { | ||
for (c <- 0 until numCols) { | ||
sb.append(leftJust(f"${darr(c * stride + r)}%.6f", 9) + " ") | ||
} | ||
sb.append("\n") | ||
} | ||
sb.toString | ||
} | ||
def printVect(dvect: DVector) = { | ||
dvect.mkString(",") | ||
} | ||
} |
Oops, something went wrong.