Titulo Estágio
Computational tools for the experimental evaluation of multiobjective optimisation algorithms
Áreas de especialidade
Sistemas Inteligentes
Local do Estágio
DEI
Enquadramento
The experimental evaluation of the performance of multiobjective optimisation algorithms is currently of great interest, especially to the evolutionary multiobjective optimisation (EMO) research community. A common scenario involves the application of a number of alternative algorithms to an appropriate set of problem instances, and the statistical analysis of the run-time and/or solution-quality results thus obtained.
Due to their stochastic nature, multiobjective evolutionary algorithms typically produce a different set of non-dominated solutions in each run according to some (unknown) probability distribution, which depends both on the problem instance and on the algorithm itself. Ultimately, the experimental evaluation of the performance of such optimisation algorithms consists of estimating and comparing aspects of the corresponding set distributions, which may also include a run-time dimension.
Objetivo
This work is concerned mainly with the development of computational tools for the experimental evaluation of multiobjective optimisation algorithms, including improved algorithms to compute the Empirical Attainment Function (EAF) and EAF-based randomisation test procedures. Common experimental settings, such as the selection of the best from a number of alternative optimisation methods or the comparison of newly developed algorithms with a state-of-the-art algorithm (the control), usually involve multiple influence factors, blocking, and multiple testing issues, but none of these are covered by current EAF-based test procedures.
Integration of such tools with existing numerical and scientific visualisation environments such as python, numpy and matplotlib or mayavi in order to make them easily available to the practitioner is another goal of this work.
Plano de Trabalhos - Semestre 1
1. Literature review on evolutionary multiobjective optimisers and performance assessment.
2. Familiarisation with the existing algorithms for the computation of the EAF and with the theoretical concepts of statistical experimental design, permutation tests and multiple comparisons, including closed test procedures.
3. Development of output-sensitive algorithms for the computation of the EAF in 3, and possibly 4, dimensions.
4. Implementation of EAF-based permutation tests for the two-sample case.
5. Evaluation of the algorithms developed.
Plano de Trabalhos - Semestre 2
6. Description of the experimental set-ups of interest and specification of the statistical test problems to be considered.
7. Implementation of the corresponding statistical hypothesis tests.
8. Comparative study involving multiple EMO algorithms and multiple optimisation problems.
9. Interpretation and discussion of the results.
10. Dissertation writing.
Condições
This work will be co-supervised by Andreia P. Guerreiro (PhD student @DEI).
The ideal candidate will have a strong background in at least two of the following areas: statistics, high-performance computing and evolutionary computation. The work will be carried out in the CISUC ECOS laboratories. A computer cluster is available to support the development and evaluation of the algorithms developed.
Observações
See https://eden.dei.uc.pt/~cmfonsec/software.html for additional information, including code and bibliography.
Orientador
Carlos Manuel Mira da Fonseca
cmfonsec@dei.uc.pt 📩