README.md

In jtilly/knitroR: R Interface for Non-Linear Constraint Optimizer Knitro

knitroR: R Package for KNITRO

Update (6/20/2016): Artelys just released their own R interface for KNITRO 10. More details here. They aptly named it "KnitroR".

knitroR provides an R interface for the commercial non-linear constraint optimizer KNITRO. This is very much work in progress. At this point, this package only brings some of the functionality from KNITRO to R. I have managed to get this package to work under Linux, Mac OS, and Windows 7. The package works with KNITRO 8 and KNITRO 9.x. Installation instructions are below.

knitroR builds an R wrapper around KNITRO's C++ interface using Rcpp.

Example

Consider the following example. The goal is to minimize a smooth function in three parameters with an inequality constraint and non-negativity constraints on all three parameters.

# load package
library("knitroR")

# define the objective function
objFun = function(x) { 
    9.0 - 8.0*x[1] - 6.0*x[2] - 4.0*x[3]+2.0*x[1]^2 + 2.0*x[2]^2 + x[3]^2 + 2.0*x[1]*x[2] + 2.0*x[1]*x[3]
}

# define the inequality constraint
c_inequality = function(x) {
    return(  x[1] + x[2] + 2.0*x[3] - 3.0  )
}

# specify non-negativity constraints
lb = c(0, 0, 0)

# define starting values
x0 = c(0.5, 0.5, 0.5)

# minimize the objective function using KNITRO
results = knitro(objFun=objFun, c_inequality = c_inequality, lb=lb, x0=x0, options="options.opt")

The code returns a list results that includes (among other things) KNITRO's exit status, the final parameters values, and the value of the objective function at the minimum.

> results$status
[1] 0
> results$x
[1] 1.3333333 0.7777778 0.4444444
> results$fval
[1] 0.1111111

For more examples see here. These examples illustrate how to pass user-defined gradients and Jacobians to KNITRO.

Documentation

• The official documentation for KNITRO is available here.
• The reference manual for knitroR is available here.

Installation

Make sure you're installing knitroR on the proper architecture. If you have the 32bit version of KNITRO, you should use a 32bit version of R and only try to build a 32bit package (i.e. turn off multiarch support using the option --no-multiarch). Similarly, if you have the 64bit version of KNITRO, you should use a 64bit version of R and only try to build a 64bit package.

Linux

To install the package under Linux you first need to set the environmental variables KNITRO, ZIENA_LICENSE, and LD_LIBRARY_PATH. This can be done directly in R (or RStudio). In the following, please adjust the paths and file names where appropriate:

Sys.setenv(KNITRO = "/path/to/knitro")
Sys.setenv(ZIENA_LICENSE = "/path/to/knitro/ziena_license.txt")
Sys.setenv(LD_LIBRARY_PATH = sprintf("%s:%s/lib", Sys.getenv("LD_LIBRARY_PATH"), Sys.getenv("KNITRO")))
source("http://jtilly.io/install_github/install_github.R")
install_github("jtilly/knitroR")

Alternatively, you can define these environmental variables using the command line:

export KNITRO=/path/to/knitro
export ZIENA_LICENSE=/path/to/knitro/ziena_license.txt
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/path/to/knitro/lib

Mac OS X

Same as under Linux with the only exception that LD_LIBRARY_PATH is called DYLD_LIBRARY_PATH: This can be done directly in R (or RStudio). In the following, please adjust the paths and file names where appropriate:

Sys.setenv(KNITRO = "/path/to/knitro")
Sys.setenv(ZIENA_LICENSE = "/path/to/knitro/ziena_license.txt")
Sys.setenv(DYLD_LIBRARY_PATH = sprintf("%s:%s/lib", Sys.getenv("DYLD_LIBRARY_PATH"), Sys.getenv("KNITRO")))
source("http://jtilly.io/install_github/install_github.R")
install_github("jtilly/knitroR")

Alternatively, you can define these environmental variables using the command line:

export KNITRO=/path/to/knitro
export ZIENA_LICENSE=/path/to/knitro/ziena_license.txt
export DYLD_LIBRARY_PATH=$DYLD_LIBRARY_PATH:/path/to/knitro/lib

Windows

I'm assuming that KNITRO was installed successfully and that all environmental variables were set appropriately. In particular, KNITRODIR must point to the directory with the KNITRO installation.

To install this package under Windows you need to download and install Rtools.

If you're using KNITRO 9.1, then you can download and install knitroR right away:

source("http://jtilly.io/install_github/install_github.R")
install_github("jtilly/knitroR")

If you're using an older version than KNITRO 9.1, you need to - Download knitroR by hand, either by cloning this repository or by downloading it from here. - Then go to src\Makevars.win in this package's source code and change the variable KNRELEASE to the appropriate version. KNRELEASE needs to be set so that it matches the name of the file knitro$(KNRELEASE).lib in your KNITRO lib directory. - Then open R and install this package by hand:

install.packages("C:\Downloads\knitroR-master", repos = NULL, type="source", INSTALL_opts="--no-multiarch")

where you need to change the path to knitroR appropriately.

Acknowledgment

Romain Francois has a KNITRO package that helped me to better understand how to get KNITRO to work in R. His package provides a deeper integration of KNITRO that allows you to register an R function as KNITRO's callback.

jtilly/knitroR documentation built on May 20, 2019, 3:13 a.m.