Software for the paper: Temporal true and surrogate fitness landscape analysis for expensive bi-objective optimisation
Creators
Description
File structure
## Folder Descriptions
- coco/: Matlab scripts using for sampling in the COCO framework.
- tsne/: Python scripts used for executing (dynamic) t_SNE analyses.
- feature-extraction/: R scripts used for extracting the landscape features based on the samples.
- performance-modelling/: R scripts used for the modelling using Random Forest models.
These scripts are used in the following paper:
Temporal true and surrogate fitness landscape analysis for expensive bi-objective optimisation
Fitness landscape analysis can provide valuable insight about key characteristics of a problem. Many real-world problems have expensive-to-compute fitness functions and are multi-objective in nature. Surrogate-assisted evolutionary algorithms are often used to tackle such problems. Despite this, literature about analysing the fitness landscapes induced by surrogate models is limited, and even non-existent for multi-objective problems. This study addresses this critical gap by comparing landscapes of the real fitness function with those of surrogate models for multi-objective functions. Moreover, it does so temporally by examining landscape features at different points in time during optimisation, in the vicinity of the population at that point in time. We consider the well-known BBOB bi-objective benchmark functions in our experiments and employ a reference-vector guided surrogate-assisted evolutionary algorithm. The results of the landscape analysis on the real fitness landscape reveal significant distinctions between features at different time points during optimisation, and between real and surrogate landscape features. Furthermore, the study demonstrates that both surrogate and real landscape features are of importance when predicting algorithm performance, and that the outcome of an algorithm can be forecast to a decent standard by sampling these during evolution. The results could help to facilitate the design of surrogate switching approaches to improve performance in multi-objective optimisation.