如何加速或矢量化 for 循环? [英] How to speed up or vectorize a for loop?

问题描述

我想通过矢量化或使用 Data.table 或其他方法来提高 for 循环的速度.我必须在 1,000,000 行上运行代码，而且我的代码真的很慢.

I would like to increase the speed of my for loop via vectorization or using Data.table or something else. I have to run the code on 1,000,000 rows and my code is really slow.

代码是不言自明的.以防万一，我在下面提供了解释.我已经包括了函数的输入和输出.希望你能帮助我更快地完成这个功能.

The code is fairly self-explanatory. I have included an explanation below just in case. I have included the input and the output of the function. Hopefully you will help me make the function faster.

我的 目标是对向量Volume"进行分箱，其中每个分箱等于 100 份.向量Volume"包含交易的股票数量.这是它的样子:

My goal is to bin the vector "Volume", where each bin is equal to 100 shares. The vector "Volume" contains the number of shares traded. Here is what it looks like:

head(Volume, n = 60)
[1]  5  3  1  5  3  1  1  1  1  1  1  1 18  1  1 18  2  7 13  2  7 13  3  2  1  1  3  2  1  1  1
[32]  1  6  6  1  1  1  1  1  1  1  1 18  2  1  1  2  1 14 18  2  1  1  2  1 14  1  1  9  5

向量binIdexVector"与Volume"等长，包含bin号；即前 100 股的每个元素获得数字 1，接下来 100 股的每个元素获得数字 2，接下来 100 股的每个元素获得数字 3，以此类推.这是向量的样子:

The vector "binIdexVector" is the same length of "Volume", and it contains the bin number; that is each element of the first 100 shares get the number 1, each elements of the next 100 shares get the number 2, each elements of the next 100 shares get the number 3, and so on. Here is what that vector looks like:

 head(binIdexVector, n = 60)
[1] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
[48] 2 2 3 3 3 3 3 3 3 3 3 3 3

这是我的功能:

#input as a vector
Volume<-c(5L, 3L, 1L, 5L, 3L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 18L, 1L, 1L, 
                   18L, 2L, 7L, 13L, 2L, 7L, 13L, 3L, 2L, 1L, 1L, 3L, 2L, 1L, 1L, 
                   1L, 1L, 6L, 6L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 18L, 2L, 1L, 
                   1L, 2L, 1L, 14L, 18L, 2L, 1L, 1L, 2L, 1L, 14L, 1L, 1L, 9L, 5L, 
                   2L, 1L, 1L, 1L, 1L, 9L, 5L, 2L, 1L, 1L, 1L, 2L, 1L, 1L, 3L, 1L, 
                   1L, 2L, 1L, 2L, 1L, 1L, 3L, 1L, 1L, 2L, 9L, 9L, 3L, 3L, 1L, 1L, 
                   1L, 1L, 5L, 5L, 8L, 8L, 2L, 1L, 2L, 1L, 10L, 10L, 10L, 10L, 10L, 
                   10L, 10L, 10L, 9L, 9L, 1L, 1L, 8L, 1L, 8L, 1L, 8L, 8L, 2L, 1L, 
                   1L, 1L, 1L, 1L, 1L, 2L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 
                   1L, 1L, 3L, 3L, 3L, 3L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 5L, 5L, 
                   1L, 2L, 7L, 1L, 2L, 7L, 1L, 1L, 1L, 1L, 2L, 1L, 10L, 1L, 1L, 
                   1L, 1L, 1L, 1L, 2L, 1L, 10L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
                   1L, 1L, 30L, 1L, 2L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 3L, 
                   1L, 2L, 2L, 1L, 1L, 1L, 1L, 2L, 2L, 1L, 2L, 2L, 2L, 1L, 2L, 1L, 
                   10L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 1L, 10L, 1L, 1L, 1L, 1L, 1L, 
                   1L, 1L, 1L, 1L, 1L, 30L, 1L, 2L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
                   1L, 1L, 3L, 1L, 2L, 2L, 1L, 1L, 1L, 1L, 2L, 2L, 1L, 2L, 2L, 2L, 
                   1L, 1L, 1L, 1L, 1L, 1L, 1L, 7L, 7L, 3L, 1L, 1L, 1L, 4L, 3L, 1L, 
                   1L, 1L, 4L, 25L, 1L, 1L, 25L, 1L, 1L, 2L, 1L, 1L, 2L, 1L, 1L, 
                   1L, 1L, 1L, 1L, 1L, 1L, 2L, 1L)

binIdexVector <- numeric(length(Volume))

# initialize 
binIdex <-1
totalVolume <-0

for(i in seq_len(length(Volume))){

  totalVolume <- totalVolume + Volume[i]  

  if (totalVolume <= 100) {

    binIdexVector[i] <- binIdex

  } else {

    binIdex <- binIdex + 1
    binIdexVector[i] <- binIdex
    totalVolume <- Volume[i]
  }
}

# output:
> dput(binIdexVector)
c(1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 
  1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 
  2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 
  3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 
  3, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 
  4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 6, 6, 6, 6, 6, 6, 
  6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 
  6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 7, 
  7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 
  7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 
  7, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 
  8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 
  8, 8, 8, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 
  9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 10, 10, 10, 10, 10, 10, 10, 
  10, 10, 10, 10, 10, 10, 10, 10, 10, 10)

非常感谢您的帮助！

> sessionInfo()
R version 3.1.2 (2014-10-31)
Platform: x86_64-w64-mingw32/x64 (64-bit)

locale:
[1] LC_COLLATE=English_United States.1252  LC_CTYPE=English_United States.1252   
[3] LC_MONETARY=English_United States.1252 LC_NUMERIC=C                          
[5] LC_TIME=English_United States.1252    

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

loaded via a namespace (and not attached):
[1] tools_3.1.2

推荐答案

当矢量化困难时，您可以使用 Rcpp.

You can use Rcpp when vectorization is difficult.

library(Rcpp)
cppFunction('
  IntegerVector bin(NumericVector Volume, int n) {
    IntegerVector binIdexVector(Volume.size());
    int binIdex = 1;
    double totalVolume =0;

    for(int i=0; i<Volume.size(); i++){
      totalVolume = totalVolume + Volume[i];
      if (totalVolume <= n) {
        binIdexVector[i] = binIdex;
      } else {
        binIdex++;
        binIdexVector[i] = binIdex;
        totalVolume = Volume[i];
      }
    }
    return binIdexVector;
  }')

all.equal(bin(Volume, 100), binIdexVector)
#[1] TRUE

它比 findInterval(cumsum(Volume), seq(0, sum(Volume), by=100)) (当然给出了不准确的答案)要快

It's faster than findInterval(cumsum(Volume), seq(0, sum(Volume), by=100)) (which of course gives an inexact answer)

这篇关于如何加速或矢量化 for 循环?的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！