An Evolution of Geometric Structures Algorithm for the Automatic Classification of HRR Radar Targets
This paper presents a novel approach to solve multiclass classification problems using pure evolutionary techniques. The proposed approach is called Evolution of Geometric Structures algorithm, and consists in the evolution of several geometric structures
- PDF / 349,834 Bytes
- 9 Pages / 430 x 660 pts Page_size
- 5 Downloads / 154 Views
stract. This paper presents a novel approach to solve multiclass classification problems using pure evolutionary techniques. The proposed approach is called Evolution of Geometric Structures algorithm, and consists in the evolution of several geometric structures such as hypercubes, hyperspheres, hyperoctahedrons, etc. to obtain a first division of the samples space, which will be re-evolved in a second step in order to solve samples belonging to two or more structures. We have applied the EGS algorithm to a well known multiclass classification problem, where our approach will be compared with several existing classification algorithms.
1
Introduction
Genetic algorithms have been widely used for solving very different problems, the majority of times related to optimization. In this field the research work has been massive in the last years and powerful algorithms based on evolutionary computation have been developed. However, there are other fields in artificial intelligence in which the genetic approach has not been so successful. In classification problems for example, the different evolutionary computing techniques have always played a secondary role: training of neural networks [2]-[7], or generation of fuzzy rules [8], but it is difficult to find a pure evolutionary technique applied to the complete resolution of the problem. The Genetic Programming technique [9] has been applied to some classification problems with some success, though its results in multiclass classification have not been so promising. The idea behind this paper is to propose a pure evolutionary technique to tackle multiclass classification problems. We have called this technique “Evolution of Geometric Structures Algorithm” (EGS), since it is based on evolving a set of geometric structures (hypercubes, hyperspheres, hyperoctahedrons, etc) to cover the samples space, and then using another evolutionary algorithm to combine them into a single classifier. As will be shown, the idea is to run a genetic algorithm encoding a set of geometric structures for each class, in such a way that the position and size of the geometric structure is evolved. A fitness H. Yin et al. (Eds.): IDEAL 2007, LNCS 4881, pp. 1151–1159, 2007. c Springer-Verlag Berlin Heidelberg 2007
1152
L. Carro-Calvo et al.
function counting the number of correct classification of samples for each class is used to guide the search. In a second step, another evolutionary algorithm is used to decide the classification of the samples within two or more geometric structures. As can be seen, only evolutionary techniques are used to solve the problem. Also, the EGS algorithm is adequate to be implemented in a parallel way, since several genetic populations must be run in the first step of the algorithm. In this paper we apply the presented technique to the automatic classification of high range resolution radar (HRR) targets, which is a hard problem of classification solved previously in the literature [11]. This kind of radar uses broad-band linear frequency modulation or step frequency waveforms to mea
Data Loading...
