nloptr的上限误差

Question

I use nloptr to model a non linear optimization problem using R.Net . I pass initial values of the parameters to the Rscript where I define lb and ub to the variables.

lb for all the variables is 0 whereas ub is "1" for some & "Inf" for some variables.

When I run the model, I am continuously getting an error as

Additional information: Error in is.nloptr(ret) : at least one element in x0 > ub

Interestingly if I provide the same initial input data from EXCEL to "R.Script", it runs fine and I never get error about x0>ub . Any insights on how to debug this or know what might be wrong? I can'f fathom why it would work with excel but not with C#.

Some details are below: Example: Parameters are in the form of Matrix as below (called my.data.var)

The constraints are such that sum of the second col and third col should be <=1 respectively. ie,

The lb for all parameters is 0 and i define it as:

lb = vector("numeric",length= length(my.data.var))

ub as:

ub.matrix <- matrix(,nrow = 2, ncol=4)  
ub.matrix
for(rownum in 1:2)  
{
  ub.matrix[rownum,1] = Inf                     
  ub.matrix[rownum,4] = Inf    
  for(colnum in 2:3) 
  {
    ub.matrix[rownum,colnum] = 1              
  }
}

The constraint func is defined as:

constraint.func <- function(my.data.var)
{
  column = 1

  constr <- vector("numeric",length = 2)   
  my.data.var.mat <- matrix(my.data.var,nrow = 2,ncol = 4,byrow = TRUE)  


  for(index in 2:3)
  {
    sum.constraint = 0
    for(prodwells in 1:2)
    {
      sum.constraint <- sum.constraint + my.data.var.mat[prodwells,index]

    }
    constr[column] <- sum.constraint - 1
    column = column+1
  }
  return(constr)

}

result <- nloptr(my.data.var,eval_f = Error.func, lb=lb,
                 ub = ub,eval_g_ineq=constraint.func,
                 opts = opts)

When I run the script, I noticed that "sum" of the first col is >1 and hence the error!

Answer 1

目前，“ Inf”限制解决了我眼前的问题。