[R] Use optim with multiple processors

[Eva Liliane Ujeneza]

Dear all,

I am trying to estimate the parameters* 3N + 2 *of a function using optim,
using a big longitudinal dataset of *N* patients observations. I defined
two functions: funct0 which calculate the residual sum of square for a
single individual, and *funct1* which calls funct0 using multiple cores and
return the total  sum of the residuals (I tried two different approaches,
both fail).

>> Code Desscription

funct1 <- (data,parameters){
test <- pbdLapply(data, funct0, parameters)   # calculate individual RSS
res <- sum(unlist(test))
return(res)
​      # Total RSS​

}

​funct1 <- (data,parameters){
jid <- data[get.jid(length(data))]
test <- lapply(data, funct0, parameters)   # calculate individual RSS
res <- sum(unlist(test))
return(res)       # Total RSS
}
​
​
result <- optim(parmeters, fn = funct1, yyy = data, method = "L-BFGS-B",
control=list(maxit=100, trace= TRUE )

<< End Code

​When I run the above with P processor​
​s, P-1 would complete​
​ (I can see the output on the screen)​
​ but the last one does not. There is no error,  and  no sign that the last
processor is still optimizing. The job just does not complete and seems
stack/still. Thus, I get no value output for *result*​. Does anyone have an
idea about what could be going wrong here ?

Cheers,

Eva


=====================================================
Eva Liliane Ujeneza, PhD student
DST/NRF Centre of Excellence in Epidemiological
Modelling and Analysis (SACEMA)

	[[alternative HTML version deleted]]