Code to evaluate an integral: translating from Matlab to R

问题

Consider the following Matlab code to approximate integrals using simulation.

function f
numSim = 1000000;
points = rand(numSim,1);
r3 = mean(feval('func3', points));

points1 = rand(numSim,1);
r8 = mean(feval('func8', points, points1));
disp([r3, r8,]);
end %f
%%%%%%%%%% Nested funcitons %%%%%%%%%%%%   
function y = func3(x)
y = exp(exp(x));
end %func3

function z = func8(x,y)
z = exp((x+y).^2);
end %func8

What I've tried in R

f <- function (func3,func8){
     numSim <- 1000000
     points <- runif(numSim)
     r3 <- mean(evaluate(points, func3))
     points1 <- runif(numSim)
     r8 <- mean(evaluate( points1,func8))
     newList<-list(r3,r8)
     return(newList)
 }
 # Nested functions  

 func3<-function(x) {
    func3 <- exp(exp(x))
    return(func3)
 }   
 func8 <- function(x,y) {
    func8<-exp((x+y)^2)
    return(func8)
 }

The first problem was a warning message:

In mean.default(evaluate(points,function)) : argument is not numeric or logical:returning NA

I added r3 <- mean(evaluate(points, func3),na.rm=TRUE)

and when I type r3 the output is [1] NA,

why is it not working correctly?

Additionally, there was a comment about -Nested functions-, I don't understand how to do that in R.

回答1:

This appears to work:

f <- function (func3,func8){
  numSim <- 1000000
  vals <- runif(numSim)  ## changed the name: 'points' is a built-in function
  r3 <- mean(sapply(vals, func3))
  vals2 <- runif(numSim)
  ## use mapply() to evaluate over multiple parameter vectors
  r8 <- mean(mapply(func8, vals, vals2))  
  newList <- list(r3,r8)
  return(newList)
}

I simplified the function definitions.

func3 <- function(x) {
  return(exp(exp(x)))
}   
func8 <- function(x,y) {
  return(exp((x+y)^2))
}

Try it out:

f(func3,func8)

I have no idea if this is correct, but I think it's a correct translation of your MATLAB code. Note that the implementation could be much faster by using vectorization: replace the sapply() and mapply() with mean(func3(vals)) and mean(func8(vals,vals2)) respectively (this only works if the functions to be evaluated are themselves appropriately vectorized, which they are in this case).

来源：https://stackoverflow.com/questions/55427848/code-to-evaluate-an-integral-translating-from-matlab-to-r

标签

matlab

function

nested

integration