0

I want to replicate this function

opt_object <- DEoptim(fn = optimFunctionDT, lower = lower, upper = upper, fnMap = mappingFun, control = DEoptim.control(trace = 500, itermax = 5000, steptol = 1000), lt_scores, ncol(lt_scores), list_matrices_def, cohort_sizes_def)

in python where mappingFun is

mappingFun <- function(x){  
  rho <- x[1]
  zvec <- tail(x, -1)  
  m <- mean(zvec)
  s <- sqrt(sum((zvec-m)^2)/(length(zvec)-1))
  zvec <- (zvec-m)/s
  c(rho, zvec)
}

I am using scipy.optimize.differential_evolution in python for replicating this function, but there is no fnMap parameter in this function. I am not able to understand where does the mappingFun go in the scipy.optimize.differential_evolution function.

This is what I have tried, but the outputs are not matching. I have used fnMap as a constraint

opt_object = scipy.optimize.differential_evolution(func = optimFunctionDT, bounds = bounds, args = (lt_scores, np.size(lt_scores,1), list_matrices, cohort_sizes), strategy = 'best1bin', maxiter = 5000, popsize = 15, tol = 0.01, mutation = (0.5, 1), recombination = 0.7, seed = 1234, callback = None, disp = False, polish = True, init = 'latinhypercube', atol = 0, updating = 'immediate', workers = 1, constraints = (nlc), x0 = None)

def mappingFun(x):
    rho = x[0]
    zvec = np.delete(x,0)
    m = np.mean(zvec)
    s = np.sqrt(sum(np.square(zvec - m)) / (len(zvec) - 1))
    zvec = (zvec - m) / s
    arr=np.insert(zvec, 0, rho)  
    return (arr)
nlc = NonlinearConstraint(mappingFun, 0, 1)
As18
  • 1
  • 1
  • Please explain what mappingFun does in R. – Andrew Nelson Dec 01 '22 at 10:00
  • Are you wanting a constraint of the parameters being within one standard deviation from the mean? – Andrew Nelson Dec 01 '22 at 10:10
  • It used to standardize the output after each population is created, but before the population is passed to the objective function. – As18 Dec 01 '22 at 13:33
  • Could you provide a MWE so we can help further? – Andrew Nelson Dec 02 '22 at 12:02
  • @AndrewNelson it's a bit difficult to provide a MWE, I just want to know is it possible to standardize the output (like calling a function and applying it on the output to standardize it) after each iteration before it goes for the next iteration for optimization. I have a list of 80 matrices which is getting optimized and gives an output list of 80 optimized values – As18 Dec 05 '22 at 09:58

0 Answers0