Algorithm configuration data mining for CMA evolution strategies

Rijn, Sander van; Wang, Hao; Stein, Bas van; Bäck, Thomas

doi:10.1145/3071178.3071205

Cited by 18 publications

(13 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, despite the ERT's current status as gold standard within this domain, its unquestioned usage across any possible setting is at least debatable. In fact, other, also multi-objective, performance measures (see, e.g., van Rijn et al, 2017;Bossek and Trautmann, 2018;Kerschke et al, 2018a) might be more applicable. And although our approach is easily transferrable to other settings (including different performance measures), changing the underlying metric likely results in a different algorithm selection model.…”

Section: Discussionmentioning

confidence: 99%

Automated Algorithm Selection on Continuous Black-Box Problems by Combining Exploratory Landscape Analysis and Machine Learning

Kerschke

Trautmann

2019

Evolutionary Computation

139

View full text Add to dashboard Cite

In this paper, we build upon previous work on designing informative and efficient Exploratory Landscape Analysis features for characterizing problems' landscapes and show their effectiveness in automatically constructing algorithm selection models in continuous black-box optimization problems. Focussing on algorithm performance results of the COCO platform of several years, we construct a representative set of high-performing complementary solvers and present an algorithm selection model that -compared to the portfolio's single best solver -on average requires less than half of the resources for solving a given problem. Therefore, there is a huge gain in efficiency compared to classical ensemble methods combined with an increased insight into problem characteristics and algorithm properties by using informative features. Acting on the assumption that the function set of the Black-Box Optimization Benchmark is representative enough for practical applications the model allows for selecting the best suited optimization algorithm within the considered set for unseen problems prior to the optimization itself based on a small sample of function evaluations. Note that such a sample can even be reused for the initial population of an evolutionary (optimization) algorithm so that even the feature costs become negligible.

show abstract

Section: Discussionmentioning

confidence: 99%

Automated Algorithm Selection on Continuous Black-Box Problems by Combining Exploratory Landscape Analysis and Machine Learning

Kerschke

Trautmann

2019

Evolutionary Computation

139

View full text Add to dashboard Cite

show abstract

“…This choice of selection allows the best offspring to be selected for the next generation and the rest of the population is deleted. (µ + λ) CMA-ES with elitism is the recent form of CMA-ES proposed in 2017 by van Rijn [8]. This method allows us to add the offspring to the original population before selecting the best fittest individuals for the next generation.…”

Section: Selection Methods In Cma-es and Dementioning

confidence: 99%

“…This method allows us to add the offspring to the original population before selecting the best fittest individuals for the next generation. From [8], it can be seen that (µ+λ) CMA-ES often performs better and faster than (µ, λ) CMA-ES.…”

Section: Selection Methods In Cma-es and Dementioning

confidence: 99%

The Effects of CMA-ES Style Selection and Restart Criteria on DE

Wineberg¹,

Opawale²

2018

mendel

View full text Add to dashboard Cite

Over the years, a lot of research has gone into the creation of different mutation operators and adaptive parameters for differential evolution (DE). However, the literature is fairly quiet about automatically setting population size and completely silent about varying the selection operator used within DE. In this paper, we steal a page from CMA-ES/IPOP: using ES-style µ+λ selection, which selects across the entire population, in place of more individualistic DE selection with its use of local selection on the target and its child. We find that the most effective choice of selection can depend on the function being optimized, although for most of the functions we tested, the original DE selection was preferable. When adding IPOP style restarting, EqualFunValHist is the most applicable of the stagnation criteria, and it is used to trigger the doubling of the population size upon restart. The initial population size is set to the same as CMA-ES. Here we find, that the restartable DE behave as well and better as regular DE with population size set as lower than the default settings used.

show abstract

“…As demonstrated within this study, by adjusting the parameters of the performance measures, users are able to adapt the ranking of a given set of algorithms, and the performance characteristics of the algorithm selectors constructed based on that set, according to their preferences -trading off, for example, running time vs. robustness. In an alternative approach, van Rijn et al (2017) utilise the advantages of two popular performance measures -ERT and fixed cost error (FCE, see, e.g. Bäck et al, 2013) -by combining and standardising them within a joint performance measure.…”

Section: Performance Measuresmentioning

confidence: 99%

Automated Algorithm Selection: Survey and Perspectives

Kerschke

Hoos

Neumann

et al. 2019

Evolutionary Computation

317

184

View full text Add to dashboard Cite

It has long been observed that for practically any computational problem that has been intensely studied, different instances are best solved using different algorithms. This is particularly pronounced for computationally hard problems, where in most cases, no single algorithm defines the state of the art; instead, there is a set of algorithms with complementary strengths. This performance complementarity can be exploited in various ways, one of which is based on the idea of selecting, from a set of given algorithms, for each problem instance to be solved the one expected to perform best. The task of automatically selecting an algorithm from a given set is known as the per-instance algorithm selection problem and has been intensely studied over the past 15 years, leading to major improvements in the state of the art in solving a growing number of discrete combinatorial problems, including propositional satisfiability and AI planning. Perinstance algorithm selection also shows much promise for boosting performance in solving continuous and mixed discrete/continuous optimisation problems.This survey provides an overview of research in automated algorithm selection, ranging from early and seminal works to recent and promising application areas. Different from earlier work, it covers applications to discrete and continuous problems, and discusses algorithm selection in context with conceptually related approaches, such as algorithm configuration, scheduling or portfolio selection. Since informative and cheaply computable problem instance features provide the basis for effective perinstance algorithm selection systems, we also provide an overview of such features for discrete and continuous problems. Finally, we provide perspectives on future work in the area and discuss a number of open research challenges.

show abstract

Algorithm configuration data mining for CMA evolution strategies

Cited by 18 publications

References 11 publications

Automated Algorithm Selection on Continuous Black-Box Problems by Combining Exploratory Landscape Analysis and Machine Learning

Automated Algorithm Selection on Continuous Black-Box Problems by Combining Exploratory Landscape Analysis and Machine Learning

The Effects of CMA-ES Style Selection and Restart Criteria on DE

Automated Algorithm Selection: Survey and Perspectives

Contact Info

Product

Resources

About