## Constructing dynamic test environments for genetic algorithms based on problem difficulty (2004)

### Cached

### Download Links

Venue: | In Proc. of the 2004 Congress on Evolutionary Computation |

Citations: | 13 - 4 self |

### BibTeX

@INPROCEEDINGS{Yang04constructingdynamic,

author = {Shengxiang Yang},

title = {Constructing dynamic test environments for genetic algorithms based on problem difficulty},

booktitle = {In Proc. of the 2004 Congress on Evolutionary Computation},

year = {2004},

pages = {1262--1269}

}

### Years of Citing Articles

### OpenURL

### Abstract

Abstract — In recent years the study of dynamic optimization problems has attracted an increasing interest from the community of genetic algorithms and researchers have developed a variety of approaches into genetic algorithms to solve these problems. In order to compare their performance an important issue is the construction of standardized dynamic test environments. Based on the concept of problem difficulty this paper proposes a new dynamic environment generator using a decomposable trap function. With this generator it is posssible to systematically construct dynamic environments with changing and bounding difficulty and hence we can test different genetic algorithms under dynamic environments with changing but controllable difficulty levels. I.

### Citations

254 |
The Design of Innovation Lessons from and for Competent Genetic Algorithms
- Goldberg
- 2002
(Show Context)
Citation Context ...y on the problem being solved but also significantly on the dynamics of environmental changes. In this paper a new dynamic environment generator is proposed based on the concept of problem difficulty =-=[12]-=-. In his recent book, Goldberg [12] remotivated and expanded upon Holand’s notation of a schema or building block (BB) [15] to understand the raw material available for genetic search. He justified th... |

243 |
A connectionist machine for genetic hillclimbing
- Ackley
- 1987
(Show Context)
Citation Context ...olution [25]. The existence of misleading BBs in deceptive functions presents serious challenge to GAs. During the early stage deceptive functions were studied and constructed based on trap functions =-=[1]-=-. Deb and Goldberg [8] have shown that for a trap function to be deceptive on average (6) F (u(⃗x)) F (u(⃗x)) b a 0 Fig. 1. 1 0.8 0 z l u(⃗x) - unitation The schematic of a generalized l-bit trap func... |

200 | The royal road for genetic algorithms: Fitness landscapes and GA performance. In Toward a practice of autonomous systems: proceeding of the first european conference on artificial life
- Mitchell, Forrest
- 1991
(Show Context)
Citation Context ...BB scalings can be constructed. In general, four types of scaling schemes have been used in test functions: 1). the uniform scaling where ci = c (c is a constant), e.g., the OneMax [1] and Royal Road =-=[17]-=- functions; 2). the exponential scaling where ci = d i−1 (d ̸= 1 is a constant), e.g., the BinInt problem; 3). the power-law scaling where ci = i d (d ̸= 0 is a constant); and 4). the mixed scaling wh... |

176 | Evolutionary Optimization in Dynamic Environments
- Branke
- 2002
(Show Context)
Citation Context ...st addressed by Goldberg and Smith [10] and has attracted a growing interest from GA’s community in recent years [2], [18]. Researchers have developed many approaches into GAs to address this problem =-=[4]-=-, such as the hypermutation scheme [5], [20], the random immigration scheme [13], memory-based methods [16], [21], and multi-population approaches [3]. In order to compare the performance of EAs with ... |

145 |
Analyzing deception in trap functions
- Deb, Goldberg
- 1993
(Show Context)
Citation Context ...tence of misleading BBs in deceptive functions presents serious challenge to GAs. During the early stage deceptive functions were studied and constructed based on trap functions [1]. Deb and Goldberg =-=[8]-=- have shown that for a trap function to be deceptive on average (6) F (u(⃗x)) F (u(⃗x)) b a 0 Fig. 1. 1 0.8 0 z l u(⃗x) - unitation The schematic of a generalized l-bit trap function. 0 0 1 2 3 4 5 6 ... |

125 |
Simple genetic algorithms and the minimal, deceptive problem
- Goldberg
- 1987
(Show Context)
Citation Context ...r of test functions have been designed. A key concern here is given to what constructs a difficult problem for GAs. Among these functions a family of fitness landscapes are called deceptive functions =-=[11]-=-, which are developed to challenge the fundamental working mechanism of GAs – Holland’s building block hypothesis [15]. The building block hypothesis suggests that GAs work by combining low-order BBs ... |

122 |
Nonstationary function optimization using genetic algorithm with dominance and diploidy
- Goldberg, Smith
- 1987
(Show Context)
Citation Context ...pace. This presents serious challenge to traditional optimization techniques as well as conventional GAs. Solving dynamic optimization problems (DOPs) by GAs was first addressed by Goldberg and Smith =-=[10]-=- and has attracted a growing interest from GA’s community in recent years [2], [18]. Researchers have developed many approaches into GAs to address this problem [4], such as the hypermutation scheme [... |

111 | Genetic algorithms for changing environments
- Grefenstette
- 1992
(Show Context)
Citation Context ...om GA’s community in recent years [2], [18]. Researchers have developed many approaches into GAs to address this problem [4], such as the hypermutation scheme [5], [20], the random immigration scheme =-=[13]-=-, memory-based methods [16], [21], and multi-population approaches [3]. In order to compare the performance of EAs with different approaches for dynamic optimization problems, on the meanwhile, resear... |

107 |
An investigation into the use of hypermutation as an adaptive operator in genetic algorithms having continuous, time-dependent nonstationary environments
- Cobb
- 1990
(Show Context)
Citation Context ...] and has attracted a growing interest from GA’s community in recent years [2], [18]. Researchers have developed many approaches into GAs to address this problem [4], such as the hypermutation scheme =-=[5]-=-, [20], the random immigration scheme [13], memory-based methods [16], [21], and multi-population approaches [3]. In order to compare the performance of EAs with different approaches for dynamic optim... |

95 | Fundamental principles of deception in genetic search,” in Foundations of Genetic Algorithms 1
- Whitley
- 1991
(Show Context)
Citation Context ...nctions where there exist low-order BBs that do not combine to form higher-order BBs: instead they form BBs resulting in a deceptive solution that is sub-optimal itself or near a sub-optimal solution =-=[25]-=-. The existence of misleading BBs in deceptive functions presents serious challenge to GAs. During the early stage deceptive functions were studied and constructed based on trap functions [1]. Deb and... |

93 | Genetic Algorithms for Tracking Changing Environments”, in Forrest (ed
- Cobb
- 1992
(Show Context)
Citation Context ...on a time varying knapsack problem where the total weight capacity of the knapsack changes over time, usually oscillating between two or more fixed values [9], [16], [18], [21]. Cobb and Grefenstette =-=[6]-=- constructed a significantly dynamic environments that switches between two predefined different fitness landscapes. The dynamic bitmatching problem [7] aims to maximize the number of bits in a string... |

66 | Nonstationary function optimization using the structured genetic algorithm
- Dasgupta, McGregor
- 1992
(Show Context)
Citation Context ... of researchers have tested their algorithms on a time varying knapsack problem where the total weight capacity of the knapsack changes over time, usually oscillating between two or more fixed values =-=[9]-=-, [16], [18], [21]. Cobb and Grefenstette [6] constructed a significantly dynamic environments that switches between two predefined different fitness landscapes. The dynamic bitmatching problem [7] ai... |

64 | A multi-population approach to dynamic optimization problems
- Branke, Kaubler, et al.
- 2000
(Show Context)
Citation Context ... many approaches into GAs to address this problem [4], such as the hypermutation scheme [5], [20], the random immigration scheme [13], memory-based methods [16], [21], and multi-population approaches =-=[3]-=-. In order to compare the performance of EAs with different approaches for dynamic optimization problems, on the meanwhile, researchers have developed several dynamic problem generators [4], [19]. Jus... |

63 | Experimental study on population-based incremental learning algorithm for dynamic optimization problems
- Yang, Yao
- 2005
(Show Context)
Citation Context ...s is related to the speed of changes (rapid or slow relative to EA time) and the severity of changes for each parameter (the step size may be large or small). 4) Revolving Fitness Landscapes In [23], =-=[24]-=-, a dynamic problem generator is proposed, which can generate dynamic environments from any binaryencoded function. Given a function f(⃗x) defined on l-bit strings (⃗x ∈ {0, 1} l ), the fitness landsc... |

60 |
A new diploid scheme and dominance change mechanism for nonstationary function optimization
- Ng, Wong
- 1995
(Show Context)
Citation Context ... [2], [18]. Researchers have developed many approaches into GAs to address this problem [4], such as the hypermutation scheme [5], [20], the random immigration scheme [13], memory-based methods [16], =-=[21]-=-, and multi-population approaches [3]. In order to compare the performance of EAs with different approaches for dynamic optimization problems, on the meanwhile, researchers have developed several dyna... |

59 | A comparison of dominance mechanisms and simple mutation on non-stationary problems
- Lewis, Hart, et al.
- 1998
(Show Context)
Citation Context ... years [2], [18]. Researchers have developed many approaches into GAs to address this problem [4], such as the hypermutation scheme [5], [20], the random immigration scheme [13], memory-based methods =-=[16]-=-, [21], and multi-population approaches [3]. In order to compare the performance of EAs with different approaches for dynamic optimization problems, on the meanwhile, researchers have developed severa... |

46 |
On the behaviour of evolutionary algorithms in dynamic environments
- Bäck
- 1998
(Show Context)
Citation Context ...s well as conventional GAs. Solving dynamic optimization problems (DOPs) by GAs was first addressed by Goldberg and Smith [10] and has attracted a growing interest from GA’s community in recent years =-=[2]-=-, [18]. Researchers have developed many approaches into GAs to address this problem [4], such as the hypermutation scheme [5], [20], the random immigration scheme [13], memory-based methods [16], [21]... |

44 |
De Jong
- Potter, A
- 1994
(Show Context)
Citation Context ...aches [3]. In order to compare the performance of EAs with different approaches for dynamic optimization problems, on the meanwhile, researchers have developed several dynamic problem generators [4], =-=[19]-=-. Just as that benchmark test problems play an important role in the study of GAs in stationary environments, constructing standardized dynamic environments plays an important role in comparing GAs fo... |

40 | Non-stationary problem optimization using the primaldual genetic algorithm
- Yang
- 2003
(Show Context)
Citation Context ...ynamics is related to the speed of changes (rapid or slow relative to EA time) and the severity of changes for each parameter (the step size may be large or small). 4) Revolving Fitness Landscapes In =-=[23]-=-, [24], a dynamic problem generator is proposed, which can generate dynamic environments from any binaryencoded function. Given a function f(⃗x) defined on l-bit strings (⃗x ∈ {0, 1} l ), the fitness ... |

36 |
Adaptation to changing environments by means of the memory based thermodynamical genetic algorithm
- Mori, Imanishi, et al.
- 1997
(Show Context)
Citation Context ...l as conventional GAs. Solving dynamic optimization problems (DOPs) by GAs was first addressed by Goldberg and Smith [10] and has attracted a growing interest from GA’s community in recent years [2], =-=[18]-=-. Researchers have developed many approaches into GAs to address this problem [4], such as the hypermutation scheme [5], [20], the random immigration scheme [13], memory-based methods [16], [21], and ... |

28 |
Evolvability in dynamic fitness landscapes: a genetic algorithm approach
- Grefenstette
- 1999
(Show Context)
Citation Context ...ep size ∆⃗xt, which can be large or small. 3) Reshaping Fitness Landscapes The third type of dynamic problem generators starts from a predefined fitness landscape, defined in n-dimensional real space =-=[14]-=-, [19], [22]. This stationary landscape is composed of a number of component landscapes (e.g., cones), each of which can change independently. Each component has its own morphology with such parameter... |

13 | An evolutionnary approach for time dependant optimization
- Collard, Escazut, et al.
- 1997
(Show Context)
Citation Context ...es [9], [16], [18], [21]. Cobb and Grefenstette [6] constructed a significantly dynamic environments that switches between two predefined different fitness landscapes. The dynamic bitmatching problem =-=[7]-=- aims to maximize the number of bits in a string that matches a given template and the template varies over time. For this type of generator, the dynamics of environmental changes is mainly characteri... |

10 |
Evolutionary Optimization in Non-Stationary Environments
- Trojanowski, Michalewicz
- 2000
(Show Context)
Citation Context ..., which can be large or small. 3) Reshaping Fitness Landscapes The third type of dynamic problem generators starts from a predefined fitness landscape, defined in n-dimensional real space [14], [19], =-=[22]-=-. This stationary landscape is composed of a number of component landscapes (e.g., cones), each of which can change independently. Each component has its own morphology with such parameters as peak he... |

5 |
De Jong (2000). Triggered Hypermutation Revisited
- Morrison, A
(Show Context)
Citation Context ... has attracted a growing interest from GA’s community in recent years [2], [18]. Researchers have developed many approaches into GAs to address this problem [4], such as the hypermutation scheme [5], =-=[20]-=-, the random immigration scheme [13], memory-based methods [16], [21], and multi-population approaches [3]. In order to compare the performance of EAs with different approaches for dynamic optimizatio... |

1 | On the behavior of evolutionary algorithms in dynamic fimess landscape - BBck - 1998 |

1 | A multipopulation approach to dynamic optimization problem - Schmidt, Schmeck - 2000 |

1 | Branke (2002). Evolutionmy Optimization in Oymic Environments - unknown authors |

1 | An evolutionary approach far time dependent optimization - Gaspar - 1997 |

1 | Nonstationary function optimization using the smchlred genetic algorithm - Dasgupla, McGregor - 1992 |

1 | The royal road for genetic algorithms: fimas landscapes and GA pfomce - Mitchell, Forrest, et al. - 1992 |

1 | Adaptation to changing environments by means of the memary based thermodynamical genetic algorithm - Mori, Kita, et al. - 1997 |

1 | De long - Morrison, A - 1999 |

1 | De long (2000). Triggered hypemutation revisited - Morrison, A - 1995 |