kNN averaging benchmarking

Paper

Trust your Neighbours: kNN-Averaging to Reduce Noise in Search-Based Approaches, Stefan Klikovits, Cédric Ho Thanh, Ahmet Cetinkaya, and Paolo Arcaini.

Structure of this repository

• make_benchmark.py: creates nmoo benchmark for synthetic problems. Use with :make_benchmark, :make_ar_benchmark or :make_gpss_benchmark
• utils.py: helper functions to generate synthetic pymoo problems.
• hv_refpoints.csv: the reference points for the hypervolume calculation of the synthetic problems
• make_casestudy.py: creates nmoo benchmark for casestudy problems. Use with :make_benchmark or :make_ar_benchmark
• src/: source code:
  • src/c_region_simulator_problem/: pymoo wrapper for the c_region_simulator problem; note that src/c_region_simulator_problem/c_region_simulator_with_pipe points to submodules/controllerTesting/controller/CRegionSimulatorWithPipe/c_region_simulator_with_pipe which is a binary you may need to recompile depending on your OS;
  • src/pendulum_cart_problem/: pymoo wrapper for the pendulum_cart problem;
• submodules/controllerTesting/: dependency for c_region_simulator and pendulum_cart;

How to run the benchmark

Please refer to the nmoo documentation for more info.

python -m nmoo run make_benchmark:make_benchmark

You might also refer to

python -m nmoo --help
python -m nmoo run --help

for more information.