## Dynamic subset selection based on a fitness case topology (2004)

### Cached

### Download Links

- [www.cs.mun.ca]
- [web.cs.mun.ca]
- [www.informatik.uni-jena.de]
- DBLP

### Other Repositories/Bibliography

Venue: | Evolutionary Computation |

Citations: | 16 - 1 self |

### BibTeX

@ARTICLE{Lasarczyk04dynamicsubset,

author = {Christian W. G. Lasarczyk and Peter Dittrich and Jena Centre For Bioinformatics and Wolfgang Banzhaf},

title = {Dynamic subset selection based on a fitness case topology},

journal = {Evolutionary Computation},

year = {2004},

volume = {12},

pages = {223--242}

}

### Years of Citing Articles

### OpenURL

### Abstract

A large training set of fitness cases can critically slow down genetic programming, if no appropriate subset selection method is applied. Such a method allows an individual to be evaluated on a smaller subset of fitness cases. In this paper we suggest a new subset selection method that takes the problem structure into account, while being problem independent at the same time. In order to achieve this, information about the problem structure is acquired during evolutionary search by creating a topology (relationship) on the set of fitness cases. The topology is induced by individuals of the evolving population. This is done by increasing the strength of the relation between two fitness cases, if an individual of the population is able to solve both of them. Our new topology–based subset selection method chooses a subset, such that fitness cases in this subset are as distantly related as is possible with respect to the induced topology. We compare topology–based selection of fitness cases with dynamic subset selection and stochastic subset sampling on four different problems. On average, runs with topology–based selection show faster progress than the others.

### Citations

3058 |
UCI repository of machine learning databases
- Blake, Merz
- 1998
(Show Context)
Citation Context ... gained from topology. 3.5 Thyroid–Problem The thyroid–problem is a real world problem. The individual’s task is to classify humans thyroid function. The dataset was obtained from the UCI–repository (=-=Blake and Merz, 1998-=-). It contains 3772 training and 3428 testing samples, each measured from one patient. A fitness case consists of a measurement vector containing 15 binary and 6 real valued entries of one human being... |

3005 |
Adaptation in natural and artificial systems
- Holland
- 1975
(Show Context)
Citation Context ...o Goldberg and Richardson (1987) ), a space formed by the knowledge of all individuals. An individual’s knowledge can spread through the population, if it leads to a higher fitness of the individual (=-=Holland, 1975-=-; Koza, 1992b). On the other hand, the lower the fitness gain is, the slower this knowledge spreads. 2.2 Topology Definition During evolution the population induces a structure on the set of fitness c... |

1333 | Binary codes capable of correcting deletions, insertions and reversals - Levenshtein - 1966 |

537 | Genetic Algorithms with sharing for multimodal function optimization - Goldberg, Richardson - 1987 |

287 |
Genetic Programming: An Introduction
- Banzhaf, Nordin, et al.
- 1998
(Show Context)
Citation Context ...1994) chooses a new subset for each generation. Each fitness case is selected independently with equal probability, which leads to varying subset sizes. Stochastic sampling (Nordin and Banzhaf, 1997; =-=Banzhaf et al., 1998-=-) chooses a new subset for each generation and for each individual, respectively, with all fitness cases having the same probability of being selected. In this article we use a third variant that we c... |

235 |
Genetic Programming: On the Programming of Computers by Natural Selection
- Koza
- 1992
(Show Context)
Citation Context ...Richardson (1987) ), a space formed by the knowledge of all individuals. An individual’s knowledge can spread through the population, if it leads to a higher fitness of the individual (Holland, 1975; =-=Koza, 1992-=-b). On the other hand, the lower the fitness gain is, the slower this knowledge spreads. 2.2 Topology Definition During evolution the population induces a structure on the set of fitness cases V . We ... |

196 | The evolution of evolvability in genetic programming
- Altenberg
- 1994
(Show Context)
Citation Context ...of high diversity and good average fitness. We showed exemplarily that runs using a topology–based selection exhibit such appropriate diversity. Wagner and Altenberg (1996) describe evolvability (see =-=Altenberg 1994-=-) as a genome’s ability to produce adaptive variants. In their opinion “adaptations are possible if improvement can be achieved in a cumulative or stepwise fashion”. As a structural key feature they c... |

181 | A study of control parameters affecting online performance of genetic algorithms for function optimization - Schaffer, Caruana, et al. - 1989 |

154 | Complex adaptations and the evolution of evolvability - WAGNER, ALTENBERG - 1996 |

137 |
Learning to tell two spirals apart
- Lang, Witbrock
- 1988
(Show Context)
Citation Context ...s to classify points belonging to one of two spirals in the x–y–plane. The spirals are intertwined and described by 97 points each. This problem has been already examined by means of neural networks (=-=Lang and Witbrock, 1989-=-) and genetic programming (Koza, 1992a). Instead of using the 194 example sized dataset, however we created an additional set including 192 examples. For this validation set we choose points next to t... |

101 | Search in a small world
- Walsh
(Show Context)
Citation Context ...led local optimum. So why are these individuals able to solve some but mostly not all fitness cases? It is well known that structure influences the efficiency of heuristics working on it (Hogg, 1996; =-=Walsh, 1999-=-). Inspired by this, we shall take a relation between the fitness cases into account, a relation detected during evolution, coded into an individual’s genotype, and spread through the population by re... |

54 | Dynamic training subset selection for supervised learning in genetic programming
- Gathercole, Ross
- 1994
(Show Context)
Citation Context ...ness calculation. These methods differ in how they choose proper subsets of the set of all fitness cases for evaluation. The simplest technique is to use a static subset. Historical subset selection (=-=Gathercole and Ross, 1994-=-), for instance, records all fitness cases that are not solved by the c○2004 (preprint) by the Massachusetts Institute of Technology Evolutionary Computation 12(2): 223-242sC. Lasarczyk, P. Dittrich, ... |

42 |
Refining the phase transition in combinatorial search
- Hogg
- 1996
(Show Context)
Citation Context ...nto a so called local optimum. So why are these individuals able to solve some but mostly not all fitness cases? It is well known that structure influences the efficiency of heuristics working on it (=-=Hogg, 1996-=-; Walsh, 1999). Inspired by this, we shall take a relation between the fitness cases into account, a relation detected during evolution, coded into an individual’s genotype, and spread through the pop... |

42 | An on-line method to evolve behavior and to control a miniature robot in real time with genetic programming
- Nordin, Banzhaf
- 1997
(Show Context)
Citation Context ...ion (Gathercole and Ross, 1994) chooses a new subset for each generation. Each fitness case is selected independently with equal probability, which leads to varying subset sizes. Stochastic sampling (=-=Nordin and Banzhaf, 1997-=-; Banzhaf et al., 1998) chooses a new subset for each generation and for each individual, respectively, with all fitness cases having the same probability of being selected. In this article we use a t... |

33 | Tsang: Partial Constraint Satisfaction Problems and Guided Local Search - Voudouris - 1996 |

25 | A genetic approach to the truck backer upper problem and the inter-twined spiral problem
- Koza
- 1992
(Show Context)
Citation Context ...Richardson (1987) ), a space formed by the knowledge of all individuals. An individual’s knowledge can spread through the population, if it leads to a higher fitness of the individual (Holland, 1975; =-=Koza, 1992-=-b). On the other hand, the lower the fitness gain is, the slower this knowledge spreads. 2.2 Topology Definition During evolution the population induces a structure on the set of fitness cases V . We ... |

22 | An Investigation of Supervised Learning in Genetic Programming
- Gathercole
- 1998
(Show Context)
Citation Context ...programs given particular fitness cases, in order to prevent a biasing influence of subset selection on evolution. Dynamic subset selection(DSS) (Gathercole and Ross, 1994; Gathercole and Ross, 1997; =-=Gathercole, 1998-=-) is a procedure based on two assumptions: (i) There is a benefit in focusing GP’s abilities on difficult fitness cases, i.e., the ones that are frequently misclassified; and (ii) there is a benefit i... |

20 | Small Populations over Many Generations can beat Large Populations over Few Generations in Genetic Programming
- Gathercole, Ross
- 1997
(Show Context)
Citation Context ...r disadvantages of certain programs given particular fitness cases, in order to prevent a biasing influence of subset selection on evolution. Dynamic subset selection(DSS) (Gathercole and Ross, 1994; =-=Gathercole and Ross, 1997-=-; Gathercole, 1998) is a procedure based on two assumptions: (i) There is a benefit in focusing GP’s abilities on difficult fitness cases, i.e., the ones that are frequently misclassified; and (ii) th... |

18 | C.: Guided Local Search - An illustrative example in function optimisation - Voudouris - 1998 |

13 | Scientific Discovery using Genetic Programming - Keijzer - 2002 |

2 |
Genetic redundancy in evolving populations of simulated robots
- Miglino, Walker
- 2002
(Show Context)
Citation Context ... approximately quadratically with training set size 2 . Therefore we recommend this method for problems where the evaluation of a fitness case is costly, such as evolving control programs for robots (=-=Miglino and Walker, 2002-=-). 2.5 Example of the Induced Structure and its Time Evolution Figure 3 shows how the structure of fitness case relations changes during evolution of a sample problem. Only edges with a weight exceedi... |

1 | Trainingsmengenselektion auf der Grundlage einer Fitnesscase–Topologie. Diploma thesis - Lasarczyk - 2002 |

1 | Evolutionary Computation Volume 12, Number 2 19 - Lasarczyk, Dittrich, et al. - 1966 |

1 |
240 Evolutionary Computation Volume 12, Number 2 Topology Based Subset Selection
- Banzhaf, Nordin, et al.
- 1998
(Show Context)
Citation Context ...1994) chooses a new subset for each generation. Each fitness case is selected independently with equal probability, which leads to varying subset sizes. Stochastic sampling (Nordin and Banzhaf, 1997; =-=Banzhaf et al., 1998-=-) chooses a new subset for each generation and for each individual, respectively, with all fitness cases having the same probability of being selected. In this article we use a third variant that we c... |