[GitHub] [spark] zhengruifeng commented on pull request #28349: [SPARK-30642][ML][PYSPARK] LinearSVC blockify input vectors

[GitBox <gi...@apache.org>]

zhengruifeng commented on pull request #28349:
URL: https://github.com/apache/spark/pull/28349#issuecomment-619509562


   dataset: [epsilon_normalized.t](https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/binary.html), numInstances=100,000, numFeatures=2,000
   
   
   testCode:
   ```
   import org.apache.spark.ml.classification._
   import org.apache.spark.storage.StorageLevel
   
   
   val df = spark.read.option("numFeatures", "2000").format("libsvm").load("/data1/Datasets/epsilon/epsilon_normalized.t").withColumn("label", (col("label")+1)/2)
   df.persist(StorageLevel.MEMORY_AND_DISK)
   df.count
   
   val svc = new LinearSVC().setBlockSize(1).setMaxIter(10)
   svc.fit(df)
   
   
   val results = Seq(1, 4, 16, 64, 256, 1024, 4096).map { size => val start = System.currentTimeMillis; val model = svc.setBlockSize(size).fit(df); val end = System.currentTimeMillis; (size, model.coefficients, end - start) }
   ```
   
   results:
   ```scala
   scala> results.map(_._2).foreach{vec => println(vec.toString.take(150))}
   [0.7494819225920858,-0.08640022531296566,2.3231273313690615,0.02563502913888478,0.041697717260290056,-0.008274929181709273,1.7654515767910142,0.073740
   [0.7494819225920862,-0.08640022531296565,2.323127331369061,0.025635029138884782,0.041697717260290056,-0.008274929181709285,1.7654515767910144,0.073740
   [0.7494819225920858,-0.08640022531296568,2.32312733136906,0.025635029138884782,0.04169771726029004,-0.008274929181709273,1.7654515767910144,0.07374022
   [0.7494819225920852,-0.08640022531296565,2.3231273313690615,0.025635029138884782,0.04169771726029004,-0.00827492918170927,1.7654515767910142,0.0737402
   [0.7494819225920866,-0.08640022531296565,2.3231273313690615,0.025635029138884782,0.041697717260290056,-0.008274929181709282,1.7654515767910144,0.07374
   [0.7494819225920862,-0.08640022531296565,2.3231273313690615,0.025635029138884866,0.04169771726029005,-0.008274929181709245,1.7654515767910144,0.073740
   [0.7494819225920862,-0.08640022531296568,2.323127331369061,0.025635029138884782,0.041697717260290056,-0.00827492918170928,1.7654515767910144,0.0737402
   
   scala> results.map(_._3)
   res13: Seq[Long] = List(43328, 8726, 8413, 9151, 9029, 8630, 9212)
   ```
   
   speed:
   blockSize=1:
   ![svc_dur_1](https://user-images.githubusercontent.com/7322292/80302195-1fd71b80-87db-11ea-9a86-1192c66b4a32.png)
   
   
   blockSize=1024
   ![svc_dur_1024](https://user-images.githubusercontent.com/7322292/80302268-8e1bde00-87db-11ea-9fc7-9794853c499e.png)
   
   
   
   
   RAM:
   blockSize=1
   ![block_svc_1](https://user-images.githubusercontent.com/7322292/80302205-31b8be80-87db-11ea-92f5-4a39ee311ca8.png)
   
   blockSize=1024
   ![block_svc_1024](https://user-images.githubusercontent.com/7322292/80302213-3ed5ad80-87db-11ea-93d3-e13467d75000.png)
   
   **Native BLAS is NOT used in this test**
   


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org