Github user jkbradley commented on a diff in the pull request:
https://github.com/apache/spark/pull/7705#discussion_r35726222
--- Diff:
mllib/src/main/scala/org/apache/spark/mllib/clustering/LDAModel.scala ---
@@ -197,8 +238,86 @@ class LocalLDAModel private[clustering] (
// TODO:
// override def topicDistributions(documents: RDD[(Long, Vector)]):
RDD[(Long, Vector)] = ???
+ /**
+ * Calculate and return per-word likelihood bound, using the `batch` of
+ * documents as evaluation corpus.
+ */
+ // TODO: calcualte logPerplexity over training set online during
training, reusing gammad instead
+ // of performing variational inference again in [[bound()]]
+ def logPerplexity(
+ batch: RDD[(Long, Vector)],
+ totalDocs: Long): Double = {
+ val corpusWords = batch
+ .flatMap { case (_, termCounts) => termCounts.toArray }
+ .reduce(_ + _)
+ val subsampleRatio = totalDocs.toDouble / batch.count()
+ val batchVariationalBound = bound(
+ batch,
+ subsampleRatio,
+ docConcentration,
+ topicConcentration,
+ topicsMatrix.toBreeze.toDenseMatrix,
+ gammaShape,
+ k,
+ vocabSize)
+ val perWordBound = batchVariationalBound / (subsampleRatio *
corpusWords)
+
+ perWordBound
+ }
+
+
+ /**
+ * Estimate the variational bound of documents from `batch`:
+ * E_q[log p(bath)] - E_q[log q(batch)]
+ */
+ private def bound(
+ batch: RDD[(Long, Vector)],
+ subsampleRatio: Double,
+ alpha: Vector,
+ eta: Double,
+ lambda: BDM[Double],
+ gammaShape: Double,
+ k: Int,
+ vocabSize: Long): Double = {
+ val brzAlpha = alpha.toBreeze.toDenseVector
+ // transpose because dirichletExpectation normalizes by row and we
need to normalize
+ // by topic (columns of lambda)
+ val Elogbeta = LDAUtils.dirichletExpectation(lambda.t).t
+
+ var score = batch.map { case (id: Long, termCounts: Vector) =>
+ var docScore = 0.0D
+ val (gammad: BDV[Double], _) =
OnlineLDAOptimizer.variationalTopicInference(
+ termCounts, exp(Elogbeta), brzAlpha, gammaShape, k)
+ val Elogthetad: BDV[Double] = LDAUtils.dirichletExpectation(gammad)
+
+ // E[log p(doc | theta, beta)]
+ termCounts.foreachActive { case (id, count) =>
--- End diff --
"id" --> "idx" to avoid variable shadowing
---
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]
