benchmark_optimization

This is for running benchmarks between optimization solvers. Enmphasis is put on making it easy to add new solvers.

To get started

1. Clone this repo
   • git clone --recurse-submodules https://github.com/Guillaume-Garrigos/benchmark_optimization.git
   • cd benchmark_optimization
   • git submodule foreach --recursive git checkout main
   • Note that we use the --recurse-submodules option here. This is for you to download all the solvers available to you.
2. [optional] Add some specific solvers in /src/solvers/
   • you can type your own solvers in a python file, eventually in some subfolder.
3. [optional] Download some datasets (the repo comes with a dummy dataset to run first experiments).
   • datasets must be placed in /datasets/, with the .txt extension.
   • datasets can be downloaded here : LibSVM.
4. Edit config.py to specify which solvers to run, with which parameters. If you want more details about what are the parameters, you can give a look at /src/config_default.py, see also the Documentation
5. Run experiments. You have different options:
   • Run python run.py or bash run.sh
   • Open a python script or notebook, and run the following lines:

   from src.benchmark import benchmark_datasets
   
   benchmark_datasets()