Constrained Optimization is a UG4 Plugin that provides various algorithms used in a constrained optimization setting.
This document should give a broad overview on ConstrainedOptimization combined with some usage examples. It also provides some brief notes concerning the installation of the plugin.
A sample problem will be added to the description.
The installation equals the standard process for adding UG4 plugins described on the ughub GitHub page. Due to the usage of OS specific functions for process generation, the plugin is not necessarily multiplatform. In its current form, the package has been shown to work on Windows 10, Linux Ubuntu and macOS Mojave 10.14.4. On Windows 10, two compilers were tested, GCC 9.3.0 and Visual Studio v16.3.10. It is assumed that the package also works for older compiler versions that support C++17. On Linux Ubuntu, only GCC was tested. On macOS Mojave the Clang compiler was used. GCC and Clang might require special flags set in the build process, which is explained below. The utilized Visual Studio compiler has not exhibited such necessities but older versions might.
Follow the steps on the ughub GitHub page. If errors occur, proceed with this text. The plugin makes use of C++ std::threads. This might necessitates activating the -pthread flag to the build process for ug4. Within your UG4 install library, go to
cd ug4/ugcore/cmake/
and open
ug_includes.cmake
Now search for
elseif("${CMAKE_CXX_COMPILER_ID}" STREQUAL "GNU")
add_cxx_flag("-Wall")
add_cxx_flag("-Wno-multichar")
add_cxx_flag("-Wno-unused-local-typedefs")
add_cxx_flag("-Wno-maybe-uninitialized")
and add
add_cxx_flag("-pthread")
to the GNU include statements. Now rebuild UG4 as described on the ughub GitHub page. The plugin should now be installed without any issues.
Follow the steps on the ughub GitHub page.. If errors occur, proceed with this text. The plugin makes use of C++11 features, like std::threads and constexpr. This might necessitates activating the -std=c++11 flag to the build process for ug4. Within your UG4 install library, go to
cd ug4/ugcore/cmake/
and open
ug_includes.cmake
Now search for
elseif("${CMAKE_CXX_COMPILER_ID}" STREQUAL "Clang")
add_cxx_flag("-Wall")
add_cxx_flag("-Wno-multichar")
and add
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11")
to the Clang include statements. This ensures that the -std=c++11 flag is only added to the C++ compiler and not the C compiler in the build process. Now rebuild UG4 as described on the ughub GitHub page.. The plugin should now be installed without any issues.
The following algorithms are currently implemented:
| Name | Description |
|---|---|
| Newton-Gauss | The Gauss Newton procedure minimizes functions that can be represented as a sum of squares. |
| Particle Swarm Optimization | The Particle Swarm Optimization (PSO) algorithm minimizes any continuous function on a bounded domain. |
| Geometry Sampler | The geometry sampler samples points from an arbitrary loss function. |
The main feature of ConstrainedOptimization is the compartmentalization of problem formulation, solution finding and underlying computation mechanism.
A problem is formulated via child instantiations of the Evaluation class type, which includes details such as the objective function to be optimized or problem specific parsing details.
Solution finding algorithms are represented as classes such as NewtonOptimizer. Computational interaction with UG4 is abstracted away in instances
of ComputationMode. Combined, this setup gives a flexible and modular platform that can be adapted for many problems at hand.
Samples are included in the documentation and the Examples folder. A thorough and detailed manual can also be found there.