#### DMCA

## Adaptive Strategy Selection in Differential Evolution (2010)

### Cached

### Download Links

Venue: | GENETIC AND EVOLUTIONARY COMPUTATION CONFERENCE (GECCO) |

Citations: | 20 - 8 self |

### Citations

773 |
Differential evolution: A simple and efficient heuristic for global optimization over continuous spaces
- Storn, Price
- 1997
(Show Context)
Citation Context ... Intelligence: Problem Solving, Control Methods, and Search; G.1.6 [Numerical Analysis]: Optimization General Terms Algorithms 1. INTRODUCTION Differential evolution (DE), proposed by Storn and Price =-=[20]-=-, is an efficient and versatile population-based, direct search algorithm for global optimization. Among DE advantages are its simple structure, ease of use, speed, and robustness, which allows its ap... |

413 |
Differential Evolution: A Practical Approach to Global Optimization
- Price, Storn, et al.
- 2005
(Show Context)
Citation Context ...ease of use, speed, and robustness, which allows its application on many real-world applications, such as data mining [1, 4], pattern recognition, digital filter design, neural network training, etc. =-=[15, 6, 3]-=-. Most recently, DE has also been used for the global permutation-based combinatorial problems [13]. Permission to make digital or hard copies of all or part of this work for personal or classroom use... |

350 | Parameter control in evolutionary algorithms
- Eiben, Hinterding, et al.
- 1999
(Show Context)
Citation Context ...it has been empirically and theoretically demonstrated that different values of parameters might be optimal at different stages of the search process (see [5, p.21] and references therein). Following =-=[5]-=-, the internal control of the parameters can be done in different ways. Deterministic methods modify the parameters values according to predefined rules; Self-Adaptive methods encode the parameters wi... |

327 | Evolutionary programming made faster
- Yao, Liu, et al.
- 1999
(Show Context)
Citation Context ...d, while the relative fitness improvement is used in our approach. 4. EXPERIMENTAL ANALYSIS In order to evaluate the performance of our approach, 13 benchmark functions (f01 − f13) were selected from =-=[26]-=- as the test suit. Functions f01 − f04 are unimodal. The Rosenbrock’s function f05 is a multi-modal function when D > 3 [18]. Function f06 is the step function, which has one minimum and is discontinu... |

159 | Problem definitions and evaluation criteria for the CEC 2005 special session on real-parameter optimization
- Suganthan, Hansen, et al.
- 2005
(Show Context)
Citation Context ...ias, we also use the same set of initial random populations to evaluate the different algorithms, as done in [11]. 4.2 Performance Criteria Four performance criteria were selected from the literature =-=[22]-=- to evaluate the performance of the algorithms. These criteria are described as follows. • Error: The error of a solution x is defined as f(x)− f(x∗), where x∗ is the global minimum of the function. T... |

107 | Differential evolution algorithm with strategy adaptation for global numerical optimization
- Qin, Huang, et al.
(Show Context)
Citation Context ...ther different strategies in [15, 21]. However, the selection of which of such strategies should be used is a difficult and crucial task for the performance of the DE, besides being problem-dependent =-=[17, 16]-=-. Some approaches have already been proposed to do this strategy selection in an autonomous way, as follows. Xie and Zhang [24] presented a swarm algorithm framework, in which a neural network is used... |

71 |
Meta-Lamarckian learning in memetic algorithms
- Ong, Keane
- 2004
(Show Context)
Citation Context ...value of each strategy, used to ensure that no operator gets lost [23]. 3.2 Credit Assignment In order to assign the credit for each strategy, we adopt the relative fitness improvement ηi proposed in =-=[12]-=- as follows: ηi = δ cfi · |pfi − cfi| (5) where i = 1, · · · , NP . δ is the fitness of the best-so-far solution in the population. pfi and cfi are the fitness of the target parent and of its offsprin... |

69 |
Self-adaptive differential evolution algorithm for numerical optimization, in: Proc
- Qin, Suganthan
(Show Context)
Citation Context ...posed a variant of DE, namely SaDE, that implements different strategies and also updates their weights in the search based on their previous success rate. They also used SaDE on constrained problems =-=[9]-=-. In [2] and [25] strategy adaptation techniques similar to SaDE are also used to enhance DE performance. To the best of our knowledge, the study on adaptive strategy selection in DE is still scarce. ... |

52 |
Accelerating Differential Evolution using an adaptive local search
- Noman, Iba
- 2008
(Show Context)
Citation Context ...riments, each function is optimized over 50 independent runs. To avoid any initialization bias, we also use the same set of initial random populations to evaluate the different algorithms, as done in =-=[11]-=-. 4.2 Performance Criteria Four performance criteria were selected from the literature [22] to evaluate the performance of the algorithms. These criteria are described as follows. • Error: The error o... |

47 |
An adaptive pursuit strategy for allocating operator probabilities
- Thierens
- 2005
(Show Context)
Citation Context ...the search. This paper is focused on the latter approach, more specifically, on the Adaptive Strategy Selection (AdapSS). Inspired by some recent works in the Genetic Algorithms community (see, e.g., =-=[23, 7]-=-), its objective is to automatically select between the available (possibly ill-known) mutation strategies, according to their performance on the current search/optimization process. To do so, there i... |

36 |
A note on the extended rosenbrock function
- Shang, Qiu
(Show Context)
Citation Context ...ance of our approach, 13 benchmark functions (f01 − f13) were selected from [26] as the test suit. Functions f01 − f04 are unimodal. The Rosenbrock’s function f05 is a multi-modal function when D > 3 =-=[18]-=-. Function f06 is the step function, which has one minimum and is discontinuous. Function f07 is a noisy quartic function. Functions f08 − f13 are multimodal functions where the number of local minima... |

31 |
Probability matching, the magnitude of reinforcement, and classifier system bidding
- Goldberg
- 1990
(Show Context)
Citation Context ...ristics and being suitable for different problems. However, choosing the best strategy for a problem at hand is a difficult task. In this work, we propose the utilization of Probability Matching (PM) =-=[8]-=- for the autonomous stratin ria -0 04 71 26 8,sv er sio ns1s- 7sA prs2 01 0 egy selection on the DE, what we refer to as PM-AdapSS-DE. The main objectives of this work are two-fold. First, the PM meth... |

27 | Self-adaptive Differential Evolution with Neighborhood Search
- Yang, Tang, et al.
- 2008
(Show Context)
Citation Context ...of DE, namely SaDE, that implements different strategies and also updates their weights in the search based on their previous success rate. They also used SaDE on constrained problems [9]. In [2] and =-=[25]-=- strategy adaptation techniques similar to SaDE are also used to enhance DE performance. To the best of our knowledge, the study on adaptive strategy selection in DE is still scarce. In order to allev... |

23 | Automatic clustering using an improved differential evolution algorithm
- Das, Abraham, et al.
- 2008
(Show Context)
Citation Context ... algorithm for global optimization. Among DE advantages are its simple structure, ease of use, speed, and robustness, which allows its application on many real-world applications, such as data mining =-=[1, 4]-=-, pattern recognition, digital filter design, neural network training, etc. [15, 6, 3]. Most recently, DE has also been used for the global permutation-based combinatorial problems [13]. Permission to... |

19 |
MODENAR: multi-objective differential evolution algorithm for mining numeric association rules
- Alatas, Akin, et al.
- 2008
(Show Context)
Citation Context ... algorithm for global optimization. Among DE advantages are its simple structure, ease of use, speed, and robustness, which allows its application on many real-world applications, such as data mining =-=[1, 4]-=-, pattern recognition, digital filter design, neural network training, etc. [15, 6, 3]. Most recently, DE has also been used for the global permutation-based combinatorial problems [13]. Permission to... |

17 | Analysis of adaptive operator selection techniques on the royal road and long k-path problems
- Fialho, Schoenauer, et al.
- 2009
(Show Context)
Citation Context ...the search. This paper is focused on the latter approach, more specifically, on the Adaptive Strategy Selection (AdapSS). Inspired by some recent works in the Genetic Algorithms community (see, e.g., =-=[23, 7]-=-), its objective is to automatically select between the available (possibly ill-known) mutation strategies, according to their performance on the current search/optimization process. To do so, there i... |

11 |
Differential Evolution: In Search of Solutions
- Feoktistov
- 2006
(Show Context)
Citation Context ...ease of use, speed, and robustness, which allows its application on many real-world applications, such as data mining [1, 4], pattern recognition, digital filter design, neural network training, etc. =-=[15, 6, 3]-=-. Most recently, DE has also been used for the global permutation-based combinatorial problems [13]. Permission to make digital or hard copies of all or part of this work for personal or classroom use... |

10 |
Advances in Differential Evolution
- Chakraborty
- 2008
(Show Context)
Citation Context ...ease of use, speed, and robustness, which allows its application on many real-world applications, such as data mining [1, 4], pattern recognition, digital filter design, neural network training, etc. =-=[15, 6, 3]-=-. Most recently, DE has also been used for the global permutation-based combinatorial problems [13]. Permission to make digital or hard copies of all or part of this work for personal or classroom use... |

10 |
2008] “Large scale global optimization using differential evolution with self-adaptation and cooperative co-evolution
- Zamuda, Brest, et al.
(Show Context)
Citation Context ...ategy selection in an autonomous way, as follows. Xie and Zhang [24] presented a swarm algorithm framework, in which a neural network is used to adaptively update the weights of the DE strategies. In =-=[27]-=-, Zamuda et al. used a fixed parameter rs to choose the strategy among three available ones. Qin et al. [17, 16] proposed a variant of DE, namely SaDE, that implements different strategies and also up... |

6 | SWAF: Swarm algorithm framework for numerical optimization
- Xie, Zhang
- 2004
(Show Context)
Citation Context ... task for the performance of the DE, besides being problem-dependent [17, 16]. Some approaches have already been proposed to do this strategy selection in an autonomous way, as follows. Xie and Zhang =-=[24]-=- presented a swarm algorithm framework, in which a neural network is used to adaptively update the weights of the DE strategies. In [27], Zamuda et al. used a fixed parameter rs to choose the strategy... |

5 |
Comparing the uni-modal scaling performance of global and local selection in a mutation-only differential evolution algorithm,” in
- Price, Ronkkonen
(Show Context)
Citation Context ...s faster at the beginning of the evolution. However, it may cause stagnation rapidly. The reason might be that the fourth strategy is based on the current solution and performs local search around it =-=[14]-=-. • From Figure 3, we can see that on most of the functions, the strategy 4 obtains the greatest probability, followed by the third, the first, and the second one. This phenomenon is reasonable. Since... |

5 |
Home page of differential evolution
- Storn, Price
- 2003
(Show Context)
Citation Context ...USA. Copyright 2010 ACM 978-1-4503-0072-8/10/07 ...$10.00. In the seminal DE algorithm [20], a single mutation strategy was used. Later on, Price and Storn suggested ten other different strategies in =-=[15, 21]-=-. However, the selection of which of such strategies should be used is a difficult and crucial task for the performance of the DE, besides being problem-dependent [17, 16]. Some approaches have alread... |

3 |
Differential Evolution: A Handbook for Global Permutation-Based Combinatorial Optimization
- Onwubolu, Davendra
- 2009
(Show Context)
Citation Context ... data mining [1, 4], pattern recognition, digital filter design, neural network training, etc. [15, 6, 3]. Most recently, DE has also been used for the global permutation-based combinatorial problems =-=[13]-=-. Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial a... |

2 |
Controlling behavioral and structural parameters in evolutionary algorithms
- Maturana, Lardeux, et al.
- 2009
(Show Context)
Citation Context ...al level of diversity is also important for the search process, and thus should also be taken into account for the rewarding of the strategies. The Compass or the Pareto-based approaches, proposed in =-=[10]-=-, could be used. Lastly, the on-line adaptation of CR and F was already shown to be beneficial for DE (e.g., in the SaDE method), and should also be considered in the near future. In addition, the sen... |

1 |
Performance comparison of adaptive and self-adaptive differential evolution algorithms
- Brest, Boscovic, et al.
- 2007
(Show Context)
Citation Context ...variant of DE, namely SaDE, that implements different strategies and also updates their weights in the search based on their previous success rate. They also used SaDE on constrained problems [9]. In =-=[2]-=- and [25] strategy adaptation techniques similar to SaDE are also used to enhance DE performance. To the best of our knowledge, the study on adaptive strategy selection in DE is still scarce. In order... |

1 |
Patients’ views of a new nurse-led continence service
- Shaw, Williams, et al.
(Show Context)
Citation Context .... The Wilcoxon signed-rank test is a non-parametric statistical hypothesis test, which can be used as an alternative to the paired t-test when the results cannot be assumed to be normally distributed =-=[19]-=-. Additionally, some representative convergence graphs are plotted in Figure 1. With respect to the quality of the final results, Table 1 indicates that our proposed PM-AdapSS-DE (with different credi... |