This paper presents a cooperative coevolutive approach for designing neural network ensembles. Cooperative coevolution is recent paradigm in evolutionary computation that allows the effective modeling of environments. Although theoretically, single with sufficient number neurons hidden layer would suffice to solve any problem, practice many real-world problems are too hard construct appropriate them. In such problems, ensembles successful alternative. Nevertheless, design complex task. this...

This paper presents COVNET, a new cooperative coevolutionary model for evolving artificial neural networks. is based on the idea of coevolving subnetworks that must cooperate to form solution specific problem, instead complete The combination this part process. best combinations be evolved together with coevolution subnetworks. Several subpopulations coevolve cooperatively and genetically isolated. individual every subpopulation are combined whole different approach from most current models...

The k -nearest neighbor ( -NN) classifier is one of the most widely used methods classification due to several interesting features, including good generalization and easy implementation. Although simple, it usually able match even outperform more sophisticated complex methods. One problems with this approach fixing appropriate value . a might be obtained using cross validation, unlikely that same could optimal for whole space spanned by training set. It evident different regions feature...

In this paper we propose a crossover operator for evolutionary algorithms with real values that is based on the statistical theory of population distributions. The theoretical distribution genes best individuals in population. proposed takes into account localization and dispersion features objective these would be inherited by offspring. Our aim optimization balance between exploration exploitation search process. order to test efficiency robustness crossover, have used set functions...

We present a new method of multiclass classification based on the combination one-vs-all and modification one-vs-one method. This methods proposed enforces strength both methods. A study behavior two identifies some sources their failure. The performance classifier can be improved if are combined in one, such way that main failure partially avoided.

In this paper, we approach the problem of constructing ensembles classifiers from point view instance selection. Instance selection is aimed at obtaining a subset instances available for training capable achieving, least, same performance as whole set. way, algorithms try to keep while reducing number in Meanwhile, boosting methods construct an ensemble iteratively focusing each new member on most difficult by means biased distribution instances. work, show how these two methodologies can be...

This paper presents a hybrid evolutionary algorithm (EA) to solve nonlinear-regression problems. Although EAs have proven their ability explore large search spaces, they are comparatively inefficient in fine tuning the solution. drawback is usually avoided by means of local optimization algorithms that applied individuals population. The use procedures called algorithms. On other hand, it well known clustering process enables creation groups (clusters) with mutually close points hopefully...

Multi-label learning is a growing field in machine research. Many applications address instances that simultaneously belong to many categories, which cannot be disregarded if optimal results are desired. Among the algorithms developed for multi-label learning, k-nearest neighbor method among most successful. However, difficult classification task, such as challenge arises approach assignment of appropriate value k. Although suitable might obtained using cross-validation, it unlikely same...

Multilabel classification as a data mining task has recently attracted increasing interest from researchers. Many current applications address problems with instances that belong to more than one class. These require the development of new, efficient methods. Advantageously using correlation among different labels can provide better performance methods manage each label separately. In recent decades, many have been developed deal multilabel datasets, which makes it difficult decide method is...

In current research, an enormous amount of information is constantly being produced, which poses a challenge for data mining algorithms. Many the problems in extremely active research areas, such as bioinformatics, security and intrusion detection, or text mining, share following two features: large sets class-imbalanced distribution samples. Although many methods have been proposed dealing with sets, most these are not scalable to very common those fields. this paper, we propose new...

Instance selection is becoming increasingly relevant due to the huge amount of data that constantly produced in many fields research. At same time, most recent pattern recognition problems involve highly complex datasets with a large number possible explanatory variables. For reasons, this abundance variables significantly harms classification or tasks. There are efficiency issues, too, because speed algorithms largely improved when complexity reduced. One approaches address have too...