PatricZhao / class

Parallel KNN (k-nearest neighbors algorithm) by OpenMP and OpenACC based on CRAN R Class package.

class

❗ This is a fork of.the CRAN R package repository. class — Functions for Classification. Homepage: http://www.stats.ox.ac.uk/pub/MASS4/

============================================================================================================== This repository used to show how to acclerate legacy code by parallel technologies including openMP and openACC.

• This is OpenMP version of KNN (k-nearest neighbors algorithm) in R.
• Modified by: Peng Zhao

---

INTALL:

library(devtools)

install_github("patriczhao/class")

---

Functionality:

train <- rbind(iris3[1:25,,1], iris3[1:25,,2], iris3[1:25,,3])
test <- rbind(iris3[26:50,,1], iris3[26:50,,2], iris3[26:50,,3])
cl <- factor(c(rep("s",25), rep("c",25), rep("v",25)))
install_github("patriczhao/class")
library(class)
set.seed(1)
ompKNN.1 <- knn(train, test, cl, k = 1)
ompKNN.3 <- knn(train, test, cl, k = 3)
ompKNN.5 <- knn(train, test, cl, k = 5)

install_github("cran/class")
library(class)
set.seed(1)
baseKNN.1 <- knn(train, test, cl, k = 1)
baseKNN.3 <- knn(train, test, cl, k = 3)
baseKNN.5 <- knn(train, test, cl, k = 5)

all.equal(baseKNN.1, ompKNN.1)
> [1] TRUE
all.equal(baseKNN.3, ompKNN.3)
> [1] TRUE
all.equal(baseKNN.5, ompKNN.5)
> [1] TRUE

---

Performance TEST:

n.cols <- 500 
n.rows <- 5000 

train  <- matrix(runif(n.cols * n.rows), nrow=n.rows)
labels <- sample(10, n.rows, replace=T) 

test.rows <- 5000  
test <- matrix(runif(n.cols * test.rows), nrow=test.rows)

install_github("patriczhao/class")
set.seed(1)
system.time(baseKNN <- knn(train, test, cl=labels, k=5, n.threads=1))
system.time(ompKNN <- knn(train, test, cl=labels, k=5, n.threads=20))

---

Result :

 original(1 -core)   : 19.4 sec
 OpenMP  (20-cores)  :  2.3 sec