Conference Paper

Evolving fuzzy classifiers using a symbiotic approach

Sharif Univ. of Technol., Tehran
DOI: 10.1109/CEC.2007.4424664 Conference: Evolutionary Computation, 2007. CEC 2007. IEEE Congress on
Source: IEEE Xplore

ABSTRACT Fuzzy rule-based classifiers are one of the famous forms of the classification systems particularly in the data mining field. Genetic algorithm is a useful technique for discovering this kind of classifiers and it has been used for this purpose in some studies. In this paper, we propose a new symbiotic evolutionary approach to find desired fuzzy rule-based classifiers. For this purpose, a symbiotic combination operator has been designed as an alternative to the recombination operator (crossover) in the genetic algorithms. In the proposed approach, the evolution starts from simple chromosomes and the structure of chromosomes gets complex gradually during the evolutionary process. Experimental results on some standard data sets show the high performance of the proposed approach compared to the other existing approaches.

0 Bookmarks
 · 
67 Views
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: In 1996 Daida et al. reviewed the case for using symbiosis as the basis for evolving complex adaptive systems [6]. Specific observations included the impact of different philosophical views taken by biologists as to what constituted a symbiotic relationship and whether symbiosis represented an operator or a state. The case was made for symbiosis as an operator. Thus, although specific cost benefit characterizations may vary, the underlying process of symbiosis is the same, supporting the operator based perspective. Symbiosis provides an additional mechanism for adaption/ complexification than available under Mendelian genetics with which Evolutionary Computation (EC) is most widely associated. In the following we review the case for symbiosis in EC. In particular, symbiosis appears to represent a much more effective mechanism for automatic hierarchical model building and therefore scaling EC methods to more difficult problem domains than through Mendelian genetics alone.
    Applications of Evolutionary Computation, EvoApplicatons 2010: EvoCOMPLEX, EvoGAMES, EvoIASP, EvoINTELLIGENCE, EvoNUM, and EvoSTOC, Istanbul, Turkey, April 7-9, 2010, Proceedings, Part I; 01/2010
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper we present a novel framework for evolving ART-based classification models, which we refer to as MOME-ART. The new training framework aims to evolve populations of ART classifiers to optimize both their classification error and their structural complexity. Towards this end, it combines the use of interacting sub-populations, some traditional elements of genetic algorithms to evolve these populations and a simulated annealing process used for solution refinement to eventually give rise to a multi-objective, memetic evolutionary framework. In order to demonstrate its capabilities, we utilize the new framework to train populations of semi-supervised Fuzzy ARTMAP and compare them with similar networks trained via the recently published MO-GART framework, which has been shown as being very effective in yielding high-quality ART-based classifiers. The experimental results show clear advantages of MOME-ART in terms of Pareto Front quality and density, as well as parsimony properties of the resulting classifiers.
    Evolutionary Computation (CEC), 2010 IEEE Congress on; 08/2010
  • [Show abstract] [Hide abstract]
    ABSTRACT: In multi-objective optimization, the two pivotal attributes of the Pareto front are density and diversity. A high-density, high-diversity Pareto front is characterized by an ample set of significantly unique solutions. With respect to classifier models, obtaining a well-sampled Pareto front maximizes the ability to identify, along the trade-o curve relating model complexity and accuracy, the best-generalizing model on a testing set. Evolutionary algorithms operating in the multi-objective space have proven, to a certain extent, successful in optimizing ART-based classifiers and approaching the Pareto-optimal set. Building on previous work, a memetic genetic algorithm has been devised that evolves in parallel subpopulations of Fuzzy ARTMAP classifier models in order to attain a Pareto front with improved density and diversity. The genetic algorithm works under a quantized objective space, wherein the population of individual models is subdivided by complexity. The incorporation of specialized structures, from the Hall of Fame to the gene pool, with systematic mechanisms, including a mutation selection scheme and simulated annealing, produces a memetic algorithm that takes advantage of global exploration and localized exploitation. Safeguards against duplication and the combined forces of global and local search manifest substantial enhancement of the density and diversity of the Pareto front. Two sets of experiments were conducted to optimize mutation and to evaluate this genetic algorithm against a simpler implementation. The first series of experiments favors a uniform selection of individuals for mutation. The second series of experiments reveals that the genetic algorithm improves upon a pre-existing multi-objective evolutionary algorithm that optimizes ART-based classifiers. The comparisons between the two algorithms were with either random or trained initialization and with di erent subpopulation sizes. Overall, the novel approach is shown to be superior in terms of hyper-area, density, and two-set coverage of the final Pareto front. Also, the champion networks produced by the novel approach exhibit greater generalization power over those of the previous work.

Full-text (2 Sources)

View
1 Download
Available from