Multi-Objective Archiving
In evolutionary multiobjective optimisation, environmental selection can be generalised as a process of taking new solutions, comparing them with the old ones, and deciding how to update the population/archive, called Archiving.
In evolutionary multiobjective optimisation, environmental selection can be generalised as a process of taking new solutions, comparing them with the old ones, and deciding how to update the population/archive, called Archiving.
- Why we need new archiving methods
[TELO21] Several artificial sequences with the simplest Pareto fronts on which state-of-the-art archiving methods fail. [Read More]
[EMO19] An empirical investigation of the optimality and monotonicity of various archiving methods. It has been found -
1) deterioration happens to all the archivers;
2) the deterioration degree can vary dramatically on different problems;
3) some archivers clearly perform better than others;
4) several popular archivers sometime return a population with majority of solutions being non-optimal (i.e. being dominated by at least one solution which was discarded previously), e.g. 34% optimal ratio for NSGA-II on FON, 41% for IBEA on DTLZ1, and 37% for MOEA/D on ZDT6.