Installing additional solvers

2024-02-15

The holiglm package uses by default the ECOS solver. However, utilizing commercial solvers such as

can significantly reduce the computation time. Although these solvers are commercial, academic licenses are available free of charge for students and academic staff. It is important to note that the source code for these solvers is not accessible from CRAN and must be obtained from the respective company’s homepage.

Internally holiglm relies on the ROI package, which provides a consistent interface across various solvers. To use the commercial solvers, the corresponding R interfaces need to be installed.

Installing MOSEK

To use the MOSEK solver in hglm, you need to install the Rmosek and ROI.plugin.mosek packages. As ROI.plugin.mosek imports Rmosek, it is recommend to install Rmosek first.

Installing Rmosek

Since Rmosek is commercial software, it must be obtained from https://www.mosek.com/. Information about installing Rmosek on Windows, MacOS and Linux can be found at https://docs.mosek.com/latest/rmosek/install-interface.html.

Installing ROI.plugin.mosek

To install ROI.plugin.mosek, you need to have the Rmosek package installed. Before installing ROI.plugin.mosek, make sure that Rmosek is correctly installed by loading it using the following command in R:

library("Rmosek")

Once Rmosek is installed, you can install ROI.plugin.mosek directly from Gitlab using the following code:

remotes::install_gitlab("roigrp/solver/ROI.plugin.mosek", INSTALL_opts = "--no-multiarch")

Since ROI.plugin.mosek only uses R code and does not link to any external library it is very unlikely that the installation will fail. If you encounter difficulties during the installation,

  1. Check if Rmosek was installed correctly, by executing library("Rmosek") in R.
  2. Check your internet connection.

Installing GUROBI

As the gurobi package is commercial software, it must be obtained from https://www.gurobi.com. The site https://www.gurobi.com/documentation provides installation instructions for GUROBI on Windows, MacOS and Linux.

Installing ROI.plugin.gurobi

To install ROI.plugin.gurobi, you need to have the gurobi package installed. Before installing ROI.plugin.gurobi, ensure that gurobi is correctly installed by loading it in R:

library("gurobi")

Once gurobi is installed, you can install ROI.plugin.gurobi directly from Gitlab using the following code:

remotes:::install_gitlab("roigrp/solver/ROI.plugin.gurobi", INSTALL_opts = "--no-multiarch")

Since ROI.plugin.gurobi only uses R code and does not link to any external library it is very unlikely that the installation will fail. If you encounter difficulties during the installation,

  1. Check if gurobi was installed correctly, by executing library("gurobi") in R.
  2. Check your internet connection.

Installing CPLEX

To use ROI.plugin.cplex:

  1. Install IBM ILOG CPLEX Optimizer
  2. Install Rcplex from CRAN.
  3. Install ROI.plugin.cplex

Installing ROI.plugin.cplex

To install ROI.plugin.cplex, you need to have the Rcplex package installed. Before installing ROI.plugin.cplex, ensure that Rcplex is correctly installed by loading it in R:

library("Rcplex")

Once Rcplex is installed, you can install ROI.plugin.cplex directly from Gitlab using the following code.

remotes:::install_gitlab("roigrp/solver/ROI.plugin.cplex", INSTALL_opts = "--no-multiarch")

Since ROI.plugin.cplex only uses R code and does not link to any external library it is very unlikely that the installation will fail. If you encounter difficulties during the installation,

  1. Check if Rcplex was installed correctly, by executing library("Rcplex") in R.
  2. Check your internet connection.