In [1] we proposed Relative Convergence Speed (RCS) as measure for the convergence speed that takes into account the properties of prediction-based optimization algorithms better than existing measures. It measures how fast an algorithm approaches the global optimum relative to the worst one.
A detailed description of RCS can be found in [1] or in this document.
[1] A. Meier, O. Kramer: "Prediction with Recurrent Neural Networks in Evolutionary Dynamic Optimization". In: K. Sim and P. Kaufmann: Applications of Evolutionary Computation (EvoApplications), 2018.
In order to run this python project you need a python installation and numpy. The code has been tested with:
- Python 3.5.2
- Ubuntu 16.04
Run the main function in rcs.py; it implements a small example.