This is a Python implementation of the Adam-based Uncrowded Hypervolume Gradient Ascent algorithm (UHV-Adam) described in:
Deist, T.M., Maree, S.C., Alderliesten, T., Bosman, P.A.N. (2020). Multi-objective Optimization by Uncrowded Hypervolume Gradient Ascent.
Link to pre-print: https://arxiv.org/abs/2007.04846
The final authenticated version of the manuscript will be available in the conference proceedings of Parallel Problem Solving from Nature – PPSN XVI.
To run the algorithm on the four quadratic benchmarks described in the manuscript (Table 2):
-
Clone the repository
-
Create a subfolder named 'statistics' within the base folder
-
Run
python3 script_run_experiment.py
Statistics of each optimization run can then be found in the subfolder 'statistics'.