Github user srowen commented on a diff in the pull request:
https://github.com/apache/spark/pull/3833#discussion_r22932026
--- Diff:
mllib/src/main/scala/org/apache/spark/mllib/classification/LogisticRegression.scala
---
@@ -61,20 +67,70 @@ class LogisticRegressionModel (
override protected def predictPoint(dataMatrix: Vector, weightMatrix:
Vector,
intercept: Double) = {
- val margin = weightMatrix.toBreeze.dot(dataMatrix.toBreeze) + intercept
- val score = 1.0 / (1.0 + math.exp(-margin))
- threshold match {
- case Some(t) => if (score > t) 1.0 else 0.0
- case None => score
+ // If dataMatrix and weightMatrix have the same dimension, it's binary
logistic regression.
+ if (dataMatrix.size == weightMatrix.size) {
+ val margin = dot(weights, dataMatrix) + intercept
+ val score = 1.0 / (1.0 + math.exp(-margin))
+ threshold match {
+ case Some(t) => if (score > t) 1.0 else 0.0
+ case None => score
+ }
+ } else {
+ val dataWithBiasSize = weightMatrix.size / (nClasses - 1)
+ val dataWithBias = if(dataWithBiasSize == dataMatrix.size) {
+ dataMatrix
+ } else {
+ assert(dataMatrix.size + 1 == dataWithBiasSize)
+ MLUtils.appendBias(dataMatrix)
+ }
+
+ val margins = Array.ofDim[Double](nClasses)
+
+ val weightsArray = weights match {
+ case dv: DenseVector => dv.values
+ case _ =>
+ throw new IllegalArgumentException(
+ s"weights only supports dense vector but got type
${weights.getClass}.")
+ }
+
+ var i = 0
+ while (i < nClasses - 1) {
--- End diff --
This may be minor but I feel like this could be more compact by creating
`margins` by `map`ping `0 until nClasses` rather than unrolling it. The outer
loop is not a tight loop here.
---
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]
