Github user MLnick commented on a diff in the pull request:
https://github.com/apache/spark/pull/18624#discussion_r127443513
--- Diff:
mllib/src/main/scala/org/apache/spark/mllib/recommendation/MatrixFactorizationModel.scala
---
@@ -286,40 +288,120 @@ object MatrixFactorizationModel extends
Loader[MatrixFactorizationModel] {
srcFeatures: RDD[(Int, Array[Double])],
dstFeatures: RDD[(Int, Array[Double])],
num: Int): RDD[(Int, Array[(Int, Double)])] = {
- val srcBlocks = blockify(srcFeatures)
- val dstBlocks = blockify(dstFeatures)
- val ratings = srcBlocks.cartesian(dstBlocks).flatMap { case (srcIter,
dstIter) =>
- val m = srcIter.size
- val n = math.min(dstIter.size, num)
- val output = new Array[(Int, (Int, Double))](m * n)
+ val srcBlocks = blockify(rank, srcFeatures).zipWithIndex()
+ val dstBlocks = blockify(rank, dstFeatures)
+ val ratings = srcBlocks.cartesian(dstBlocks).map {
+ case (((srcIds, srcFactors), index), (dstIds, dstFactors)) =>
+ val m = srcIds.length
+ val n = dstIds.length
+ val dstIdMatrix = new Array[Int](m * num)
+ val scoreMatrix = Array.fill[Double](m *
num)(Double.NegativeInfinity)
+ val pq = new BoundedPriorityQueue[(Int,
Double)](num)(Ordering.by(_._2))
+
+ val ratings = srcFactors.transpose.multiply(dstFactors)
+ var i = 0
+ var j = 0
+ while (i < m) {
+ var k = 0
+ while (k < n) {
+ pq += dstIds(k) -> ratings(i, k)
+ k += 1
+ }
+ var size = pq.size
+ while (size > 0) {
+ size -= 1
+ val factor = pq.poll()
--- End diff --
Is it really necessary to add `poll`? For size of `k` (which is usually
very small), the approach of `pq.foreach` should suffice and is simpler
---
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 [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
