Benchmarking a hybrid multi level single linkagealgorithm on the bbob noiseless testbed

Abstract

Multi Level Single Linkage (MLSL) is a well known stochastic global optimization method. In this paper, a new hybrid variant (HMLSL) of the MLSL algorithm is presented. The most important improvements are related to the sampling phase: the sample is generated from a Sobol quasi-random sequence and a few percent of the population is further improved by using crossover and mutation operators like in a traditional differential evolution (DE) method.The aim of this study is to evaluate the performance of the new HMLSL algorithm on the testbed of 24 noiseless functions. The new algorithm is also compared against a simple MLSL and a traditional DE in order to identify the benefits of the applied improvements.The results confirm that the HMLSL outperforms the MLSL and DE methods. The new method has a larger probability of success and usually is faster especially in the final stage of the optimization than the other two algorithms.

Personalizing performance regression models to black-box optimization problems

“…The raw regression data has been collected using Python implementation, while the personalized approach and all the evaluations have been performed using R. Our fixed-budget regression is inspired by [21], but applied here to the more diverse set of algorithms suggested in [24]. Concretely, we aim at predicting the performance of the following 12 algorithms: BrentSTEPqi [36], BrentSTEPrr [36], CMA-ES-CSA [1], HCMA [26], HMLSL [33], IPOP400D [2], MCS [19], MLSL [33], OQNLP [34], fmincon [34], fminunc [34], and BIPOP-CMA-ES [13]. Note here that the latter does not appear in the portfolio analyzed in [24], but it was added since for one algorithm from the original study the raw performance data was missing.…”

Towards dynamic algorithm selection for numerical black-box optimization

“…While some algorithms occur in both Figures 6 and 7, many are included only once, indicating that they are relatively good choices for one part of the search, but not the remainder. e clearest example of this is HMLSL [30], which performs very well as A 1 , but has relatively high I 2 -values. is is caused by the fact that this algorithm typically converges quickly to a value close to the optimum, but has issues in the nal exploitation phase, thus only being bene cial to use at the start of the search.…”

“…To show that dynamic algorithm selection is also applicable to smaller portfolio's, we limit ourselves to 5 algorithms. ese are representative of some widely used algorithm families: Nelder-Doerr [8], DE-Auto [40], Bipop-aCMA-Step [22], HMLSL [30] and PSO-BFGS [41].With this reduced algorithm portfolio, we can study the improvements over their respective VBS static in more detail, and nd interesting algorithms combinations to explore further.…”

The impact of hyper-parameter tuning for landscape-aware performance regression and algorithm selection

“…The algorithm portfolio we chose for this work was suggested in [33] for its diversity. It consists of the following 12 algorithms: BrentSTEPqi [48], BrentSTEPrr [48], CMA-ES-CSA [1], HCMA [36], HMLSL [45], IPOP400D [2], MCS [23], MLSL [45], OQNLP [46], fmincon [46], fminunc [46], and BIPOP-CMA-ES [16]. Note that, due to the unavailability of the raw performance data for one of the algorithms in the original study, the BIPOP-CMA-ES was added instead of the missing one.…”