Uploaded image for project: 'Spark'
  1. Spark
  2. SPARK-30178

RobustScaler support bigger numFeatures

    XMLWordPrintableJSON

    Details

    • Type: Improvement
    • Status: Resolved
    • Priority: Minor
    • Resolution: Fixed
    • Affects Version/s: 3.0.0
    • Fix Version/s: 3.0.0
    • Component/s: ML
    • Labels:
      None

      Description

      It is a bottleneck to collect the whole Array[QuantileSummaries] from executors,

      since a QuantileSummaries is a large object, which maintains large arrays of size 10k(defaultCompressThreshold)/50k(defaultHeadSize).

      So we need to compute the ranges/medians more distributedly.

      In Spark-Shell with default params, I processed dataset with numFeatures=69,200, and current impl fail due to OOM.

        Attachments

          Issue Links

            Activity

              People

              • Assignee:
                podongfeng zhengruifeng
                Reporter:
                podongfeng zhengruifeng
              • Votes:
                0 Vote for this issue
                Watchers:
                1 Start watching this issue

                Dates

                • Created:
                  Updated:
                  Resolved: