Name
Title | ||
|---|---|---|
A Preliminary Study of Autoencoding Evolutionary Search with Selection of Problem Domains |
Abstract | ||
|---|---|---|
In the last decades, evolutionary search has been extensively studied in the literature, and successfully applied in many real-world applications for solving complex optimization problems. Traditional evolutionary search methods generally start the search from scratch, and ignore the relationship between the current problem of interests and past solved problems. However, problems seldom exist in isolation. Recently, the research topic of enhancing the search performance of evolutionary search by transferring knowledge from past solved related problems has attracted many attentions. In particular, the autoencoding evolutionary search paradigm is one of the latest approaches proposed to leverage the useful knowledge embedded in past search experiences for enhanced optimization performance. The efficacy of the autoencoding evolutionary search has been confirmed on both benchmarks and real-world applications. However, as there is no problem selection process, the original autoencoding evolutionary search transfers knowledge from all the available past problems without the consideration of problem relationships, which could cause negative transfer. It is also computational expensive if there are huge number of past problems. Taking this cue, in this paper, we embark a preliminary study to improve the autoencoding evolutionary search by proposing a similarity measure between problems based on the Spearman's rank correlation coefficient. Empirical studies with commonly used evolutionary solvers on benchmark problems are presented to verify the effectiveness of the proposed method. |
Year | DOI | Venue |
|---|---|---|
2018 | 10.1109/CEC.2018.8477641 | 2018 IEEE Congress on Evolutionary Computation (CEC) |
Keywords | Field | DocType |
problem relationships,complex optimization problems,autoencoding evolutionary search paradigm,problem selection process,evolutionary solvers,similarity measure,Spearman's rank correlation coefficient | Negative transfer,Similarity measure,Computer science,Artificial intelligence,Optimization problem,Empirical research,Machine learning,Genetic algorithm,Benchmark (computing) | Conference |
ISBN | Citations | PageRank |
978-1-5090-6018-4 | 0 | 0.34 |
References | Authors | |
12 | 5 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Ruoting Mal | 1 | 0 | 0.34 |
| Lei Zhou | 2 | 118 | 28.02 |
| Kai Liu | 3 | 162 | 19.16 |
| Chao Chen | 4 | 2032 | 185.26 |
| Xuefeng Xie | 5 | 3 | 1.54 |