I have two RDDs with this structure
org.apache.spark.rdd.RDD[(Long, org.apache.spark.mllib.linalg.Vector)]
Here each row of RDD contains an index Long
and a vector org.apache.spark.mllib.linalg.Vector
. I want to add each component of the Vector
into the corresponding component of other Vector
present in a row of other RDD. Each vector of first RDD should be added to each vector of other RDD.
An example would look like this:
RDD1:
Array[(Long, org.apache.spark.mllib.linalg.Vector)] =
Array((0,[0.1,0.2]),(1,[0.3,0.4]))
RDD2:
Array[(Long, org.apache.spark.mllib.linalg.Vector)] =
Array((0,[0.3,0.8]),(1,[0.2,0.7]))
Result:
Array[(Long, org.apache.spark.mllib.linalg.Vector)] =
Array((0,[0.4,1.0]),(0,[0.3,0.9]),(1,[0.6,1.2]),(1,[0.5,1.1]))
Please consider the same situation using List instead of Array.
Here is my solution:
val l1 = List((0,List(0.1,0.2)),(1,List(0.1,0.2)))
val l2 = List((0,List(0.3,0.8)),(1,List(0.2,0.7)))
var sms = (l1 zip l2).map{ case (m, a) => (m._1, (m._2, a._2).zipped.map(_+_))}
Let's experiment with Array :)
Instead of driver code you can do all this in transformation . This will be helpful if you have large rdds. This will perform less shuffling too.
val a:RDD[(Long, org.apache.spark.mllib.linalg.Vector)]= sc.parallelize(Array((0l,Vectors.dense(0.1,0.2)),(1l,Vectors.dense(0.3,0.4))))
val b:RDD[(Long, org.apache.spark.mllib.linalg.Vector)]= sc.parallelize(Array((0l,Vectors.dense(0.3,0.8)),(1l,Vectors.dense(0.2,0.7))))
val ab= a join b
val result=ab.map(x => (x._1,Vectors.dense(x._2._1.apply(0)+x._2._2.apply(0),x._2._1.apply(1)+x._2._2.apply(1))))
The technical post webpages of this site follow the CC BY-SA 4.0 protocol. If you need to reprint, please indicate the site URL or the original address.Any question please contact:yoyou2525@163.com.