## Fast Evolutionary Programming (1996)

Venue: | Proceedings of the Fifth Annual Conference on Evolutionary Programming |

Citations: | 35 - 4 self |

### BibTeX

@INPROCEEDINGS{Yao96fastevolutionary,

author = {Xin Yao and Yong Liu},

title = {Fast Evolutionary Programming},

booktitle = {Proceedings of the Fifth Annual Conference on Evolutionary Programming},

year = {1996},

pages = {451--460},

publisher = {MIT Press}

}

### Years of Citing Articles

### OpenURL

### Abstract

Evolutionary programming (EP) has been applied to many numerical and combinatorial optimisation problems successfully in recent years. One disadvantage of EP is its slow convergence to a good near optimum for some function optimisation problems. In this paper, we propose a fast EP (FEP) which uses a Cauchy instead of Gaussian mutation operator as the primary search operator. The relationship between FEP and classical EP (CEP) is similar to that between the fast simulated annealing and the classical version. Extensive empirical studies have been carried out to evaluate the performance of FEP for different function optimisation problems. Fifty runs have been conducted for each of the 23 test functions in our studies. Our experimental results show that FEP performs much better than CEP for multi-modal functions with many local minima while being comparable to CEP in performance for unimodal and multi-modal functions with only a few local minima. We emphasise in the paper that no single al...

### Citations

1451 |
An Introduction to Probability Theory
- Feller
- 1966
(Show Context)
Citation Context ...t in CEP except for the mutation operator. The one-dimensional Cauchy density function centred at the origin is defined by: f t (x) = 1 t t 2 + x 2 ; \Gamma1 ! x ! 1; where t ? 0 is a scale parameter =-=[10]-=-(pp.51). The corresponding distribution function is F t (x) = 1 2 + 1 arctan i x t j : The shape of f t (x) resembles that of the Gaussian density function but approaches the axis so slowly that an ex... |

640 |
Artificial intelligence through simulated evolution
- FOGEL, OWENS, et al.
- 1966
(Show Context)
Citation Context ...rithm and a class of problems which are most amenable to the algorithm. 1 Introduction Although evolutionary programming (EP) was first proposed as an evolutionary approach to artificial intelligence =-=[1]-=-, it has been recently applied to many numerical and combinatorial optimisation problems successfully [2, 3, 4]. Optimisation by EP can be summarised into two major steps: 1. Mutate all the solutions ... |

472 |
Evolution and Optimum Seeking
- Schwefel
- 1995
(Show Context)
Citation Context ... FEP's performance if a different mutation scheme for j is used in FEP since Eq.(2) is designed to work with the Gaussian not Cauchy mutation operator. 4 Test Functions We use 23 well-known functions =-=[2, 12, 13, 9]-=- in our experimental studies. This number is larger than the number used in most other papers. However, we believe it is necessary. Our purpose is not to show FEP is better or worse than CEP, but to f... |

326 |
An overview of evolutionary algorithms for parameter optimization
- Back, Schwefel
- 1993
(Show Context)
Citation Context ... as follows. Section 2 describes the global minimisation problem considered in this paper and the CEP used to solve them. The CEP algorithm given follows suggestions from Fogel [3, 8] and Back et al. =-=[9]-=-. Section 3 describes the FEP and its implementation. Section 4 gives the 23 functions used in our studies. Section 5 presents our experimental results and discussions. Finally, Section 6 concludes wi... |

323 | Evolving neural networks through augmenting topologies
- Stanley, Miikkulainen
- 2002
(Show Context)
Citation Context ...al is to find a neural network that satisfies a given task. Evolutionary programming is used to evolve neural network topologies and weights. The encoding and mutation operators are adopted from NEAT =-=[7]-=-. Both serial and parallel EP are implemented. We compare the performance of the serial and parallel version by using the problem of building an XOR network. Like NEAT, the population starts minimally... |

242 | No free lunch theorems for search
- Wolpert, Macready
- 1995
(Show Context)
Citation Context ...However, we believe it is necessary. Our purpose is not to show FEP is better or worse than CEP, but to find out when FEP is better (or worse) than CEP for what kind of problems. Wolpert and Macready =-=[14]-=- have proved that no single search algorithm is best on average for all problems. If the number of test problems is small, it would be very difficult to make a generalised conclusion. It also has the ... |

210 | Evolutionary programming made faster
- Yao, Liu, et al.
- 1999
(Show Context)
Citation Context ...on on EP. The first part will show the concept and the implementation of parallel EP. Both serial and parallel EP in this part are based on the Improved Fast Evolutionary Programming (IFEP) algorithm =-=[3]-=-. We conduct some experiments in order to compare the quality of solutions and the execution time between the serial version and the parallel version. In the second part, we will use the problem of bu... |

182 |
An introduction to simulated evolutionary optimization
- Fogel
- 1994
(Show Context)
Citation Context ...this paper is organised as follows. Section 2 describes the global minimisation problem considered in this paper and the CEP used to solve them. The CEP algorithm given follows suggestions from Fogel =-=[3, 8]-=- and Back et al. [9]. Section 3 describes the FEP and its implementation. Section 4 gives the 23 functions used in our studies. Section 5 presents our experimental results and discussions. Finally, Se... |

155 | A new evolutionary system for evolving artificial neural networks
- Yao, Liu
- 1997
(Show Context)
Citation Context ...s normally less than the population size in the serial version (µ). Several parallel implementations for EP have been proposed. Riessen et al. [4] applied a farmer/worker model to the EPNet algorithm =-=[5]-=-. In their implementation, the farmer node generated a global population, and then divided the population into several subpopulations. Each worker node was assigned a subpopulation, and then performed... |

147 | A survey of parallel genetic algorithms
- Cantu-Paz
- 1997
(Show Context)
Citation Context ...roach can be applied in numerous ways and this leads to different models. Genetic algorithms (GAs) are the most popular search algorithms that have been used in the studies of parallel EAs. Cantú-Paz =-=[1]-=- and Alba and Troya [2] have surveyed parallel GAs. Several techniques used in these studies are not limited to parallel GAs. They can be applied to other classes of EAs, such as genetic programming (... |

62 |
Evolving Artificial Intelligence
- Fogel
- 1992
(Show Context)
Citation Context ...ary programming (EP) was first proposed as an evolutionary approach to artificial intelligence [1], it has been recently applied to many numerical and combinatorial optimisation problems successfully =-=[2, 3, 4]-=-. Optimisation by EP can be summarised into two major steps: 1. Mutate all the solutions in the current population, and 2. Select the next generation from the mutated and the current solutions. These ... |

48 |
A Survey of Parallel Distributed Genetic Algorithms
- Alba, Troya
- 1999
(Show Context)
Citation Context ... numerous ways and this leads to different models. Genetic algorithms (GAs) are the most popular search algorithms that have been used in the studies of parallel EAs. Cantú-Paz [1] and Alba and Troya =-=[2]-=- have surveyed parallel GAs. Several techniques used in these studies are not limited to parallel GAs. They can be applied to other classes of EAs, such as genetic programming (GP) and evolutionary pr... |

47 |
System Identification through Simulated Evolution: A Machine Learning Approach to Modeling
- Fogel
- 1991
(Show Context)
Citation Context ...ary programming (EP) was first proposed as an evolutionary approach to artificial intelligence [1], it has been recently applied to many numerical and combinatorial optimisation problems successfully =-=[2, 3, 4]-=-. Optimisation by EP can be summarised into two major steps: 1. Mutate all the solutions in the current population, and 2. Select the next generation from the mutated and the current solutions. These ... |

44 | Gradual distributed real-coded genetic algorithms
- Herrera, Lozano
(Show Context)
Citation Context ... make a fair comparison, we will let P2 creates two offspring in each mutation and selects the best one, like the P1 algorithm. The use of parallel multiresolution search is not new. Some researchers =-=[6]-=- have already investigated this technique in a parallel GA. In this paper, we will examine this idea in EP. 4) Parallel IFEP without migration (P3) : This algorithm is identical to the P1 algorithm, e... |

42 | Applying evolutionary programming to selected travelling salesman problems
- Fogel
- 1993
(Show Context)
Citation Context ...ary programming (EP) was first proposed as an evolutionary approach to artificial intelligence [1], it has been recently applied to many numerical and combinatorial optimisation problems successfully =-=[2, 3, 4]-=-. Optimisation by EP can be summarised into two major steps: 1. Mutate all the solutions in the current population, and 2. Select the next generation from the mutated and the current solutions. These ... |

25 |
Nonconvex optimization by fast simulated annealing
- Szu, Hartley
- 1987
(Show Context)
Citation Context ...cribe the new EP as fast EP (FEP) while the EP proposed by Fogel [2] as the classical EP (CEP) in this paper. The motivation of introducing Cauchy mutation into EP comes from fast simulated annealing =-=[5, 6, 7]-=-. We have carried out many empirical studies of both FEP and CEP in order to evaluate the relative strength and weakness of FEP and CEP for different problems. Such results show that Cauchy mutation i... |

21 | Simulated annealing with extended neighbourhood
- Yao
- 1991
(Show Context)
Citation Context ...cribe the new EP as fast EP (FEP) while the EP proposed by Fogel [2] as the classical EP (CEP) in this paper. The motivation of introducing Cauchy mutation into EP comes from fast simulated annealing =-=[5, 6, 7]-=-. We have carried out many empirical studies of both FEP and CEP in order to evaluate the relative strength and weakness of FEP and CEP for different problems. Such results show that Cauchy mutation i... |

16 |
A new simulated annealing algorithm
- Yao
- 1995
(Show Context)
Citation Context ...cribe the new EP as fast EP (FEP) while the EP proposed by Fogel [2] as the classical EP (CEP) in this paper. The motivation of introducing Cauchy mutation into EP comes from fast simulated annealing =-=[5, 6, 7]-=-. We have carried out many empirical studies of both FEP and CEP in order to evaluate the relative strength and weakness of FEP and CEP for different problems. Such results show that Cauchy mutation i... |

6 |
Improving monte carlo efficiency by increasing variance
- Fishman, Kulkarni
- 1992
(Show Context)
Citation Context ...x) resembles that of the Gaussian density function but approaches the axis so slowly that an expectation does not exist. As a result, the variance of the Cauchy distribution is infinite. Some studies =-=[11]-=- have indicated the benefit of increasing the variance in MonteCarlo algorithms, which can be regarded as a type of generate-and-test algorithms. Fast simulated annealing is an example where an increa... |

6 |
Global Optimisation
- Törn, Zilinskas
- 1989
(Show Context)
Citation Context ... FEP's performance if a different mutation scheme for j is used in FEP since Eq.(2) is designed to work with the Gaussian not Cauchy mutation operator. 4 Test Functions We use 23 well-known functions =-=[2, 12, 13, 9]-=- in our experimental studies. This number is larger than the number used in most other papers. However, we believe it is necessary. Our purpose is not to show FEP is better or worse than CEP, but to f... |

4 | Pepnet: parallel evolutionary programming for constructing artificial neural networks
- Riessen, Williams, et al.
- 1997
(Show Context)
Citation Context ...population size in each node of the parallel version (a) is normally less than the population size in the serial version (!4 Several parallel implementations for EP have been proposed. Riessen er al. =-=[4]-=- applied a farmerlworker model tt3 the EPNet algorithm [5]. In their implementation, the farmer node generated a global population, and then divided the popuhtion into several subpopulations. Each wor... |

1 | A survey of parallel genetic algorithms. Calculateun Porolllles, Reseau el Sjvtem Repom's - Canhi-Paz - 1998 |

1 |
Troya.:' A slirvey of parallel distributed genetic algorithms
- Alba, M
- 1999
(Show Context)
Citation Context ... numerous ways and this leads to different models. Genetic algorithms (GAS) are the most popular search algorithms that have been used in the studies of parallel EAs. Cantc-Paz 111 and Alba and Troya =-=[2]-=- have surveyed parallel GAS. Several techniques used in these studies are not limited to parallel GAS. They can be applied to other classes of EAs, such as genetic programming (GP) and evolutionary pr... |

1 |
Eyolutioaary programming made faster
- Lin
- 1999
(Show Context)
Citation Context ...on on EP. The first part will show the concept and the implementation of parallel EP. Both serial and parallel EP in this part are based on the Improved Fast Evolutionary Programming (IFEP) algorithm =-=[3]-=-. We conduct some experiments in order to compare the quality of solutions and the execution time between the serial version and the parallel version. In the second part, we will use the problem of bu... |