A new hybrid methodology for cooperative-coevolutionary optimization of radial basis function networks

Rivera, Antonio J.; Rojas, Ignacio; Ortega, Julio; Jesus, María José del

doi:10.1007/s00500-006-0128-9

Cited by 23 publications

(12 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The models correspond with the most well-known and employed models in each methodology, such as evolutionary feature and instance selection (Cano et al 2003;Llorà and Garrell 2003), evolutionary fuzzy rule learning and Mamdani rule tuning (Alcalá et al 2006;del Jesus et al 2004;Otero and Sánchez 2006), genetic artificial neural networks (Martínez-Estudillo et al 2006;Rivera et al 2007 With all of these function blocks, we can affirm that KEEL can be useful by different types of user, each of whom may expect to find specific features in a DM software.…”

Section: Keelmentioning

confidence: 99%

KEEL: a software tool to assess evolutionary algorithms for data mining problems

et al. 2008

Self Cite

View full text Add to dashboard Cite

This paper introduces a software tool named KEEL which is a software tool to assess evolutionary algorithms for Data Mining problems of various kinds including as regression, classification, unsupervised learning, etc. It includes evolutionary learning algorithms based on different approaches: Pittsburgh, Michigan and IRL, as well as the integration of evolutionary learning techniques with different pre-processing techniques, allowing it to perform a complete analysis of any learning model in comparison to existing software tools. Moreover, KEEL has been designed with a double goal: research and educational.

show abstract

Section: Keelmentioning

confidence: 99%

KEEL: a software tool to assess evolutionary algorithms for data mining problems

et al. 2008

Self Cite

View full text Add to dashboard Cite

show abstract

“…• The number of network outputs in Rivera et al (2007) was set to one because it solved regression problem. In CO 2 RBFN, the network has one output for each class in the addressed dataset.…”

Section: Replacement Strategymentioning

confidence: 99%

“…• New and simpler credit assignment parameters to measure the contribution and the error are defined in CO 2 RBFN in order to increase the efficiency and to only penalize RBFs which represent a small set of patterns. • In Rivera et al (2007) operators were applied to the worst RBFs, but in CO 2 RBFN, operators are applied to the whole population. Therefore, a high level of exploration-exploitation is promoted.…”

Section: Replacement Strategymentioning

confidence: 99%

“…Therefore, a high level of exploration-exploitation is promoted. • Regarding the operators, in Rivera et al (2007), two operators were considered: an RBF deleting operator, and a biased mutation operator based on gradient error techniques for tuning the RBF centre and radius. CO 2 RBFN considers four operators: an RBF deleting operator; a random mutation operator which modifies RBF width and centre and so improves the exploration features; a new biased operator which carries out a more specialized analysis of the local environment information, based on clustering techniques, so the exploitation characteristics are improved; finally, in CO 2 RBFN, a new operator that maintains the RBF parameters is introduced in order to achieve a parsimonious evolution of the population.…”

Section: Replacement Strategymentioning

confidence: 99%

See 1 more Smart Citation

CO2RBFN: an evolutionary cooperative–competitive RBFN design algorithm for classification problems

et al. 2009

View full text Add to dashboard Cite

This paper presents a new evolutionary cooperative-competitive algorithm for the design of Radial Basis Function Networks (RBFNs) for classification problems. The algorithm, CO 2 RBFN, promotes a cooperativecompetitive environment where each individual represents a radial basis function (RBF) and the entire population is responsible for the final solution. The proposal considers, in order to measure the credit assignment of an individual, three factors: contribution to the output of the complete RBFN, local error and overlapping. In addition, to decide the operators' application probability over an RBF, the algorithm uses a Fuzzy Rule Based System. It must be highlighted that the evolutionary algorithm considers a distance measure which deals, without loss of information, with differences between nominal features which are very usual in classification problems. The precision and complexity of the network obtained by the algorithm are compared with those obtained by different soft computing methods through statistical tests. This study shows that CO 2 RBFN obtains RBFNs with an appropriate balance between accuracy and simplicity, outperforming the other methods considered.

show abstract

“…A recent reference to topology design can be found in Rivera et al (2007) where a multiobjective cooperative coevolutive hybrid algorithm for the design of a Radial Basis Function Network is presented. In the case of weights, a method based in path-relinking heuristics has been recently proposed to train a single-layer feed-forward neural network (El-Fallahi et al 2006).…”

mentioning

confidence: 99%

Evolutionary and metaheuristics based data mining

2008

View full text Add to dashboard Cite

In a strict sense Data Mining (DM) is only one step of the Knowledge Discovery from Databases (KDD) process (Fayyad et al. 1996), concretely the phase consisting of the application of specific algorithms to extract patterns from data. However, the term Data Mining has been popularly used as a synonymous of KDD or at least in a wider sense that comprise the phases of data pre-processing and pattern/model discovering (or DM).DM poses a great range of interesting (NP-hard) problems that consists mainly in searching for: the pattern/ model that best describe the data, the more predictive subset of variables, the more accurate parameter configuration, etc. Because of this, and of the success obtained by metaheuristics when applied to other combinatorial/ numerical optimization problems, metaheuristics have been widely applied to solve DM problems during the last years. In fact, the use of evolutionary algorithms and metaheuristics in general to approach DM-based problems (EMBDM) is a hot topic of research nowadays.Two particular features of DM-based optimization problems make them a challenge for the application of metaheuristics algorithms: (1) apart of being NP-hard problems, the instances that we want to solve are growing in size every day, e.g., size of the databases, number of variables involved, complexity of the discovered model, etc.; and (2) the evaluation function is computationally very expensive because it usually requires to transform the obtained individual into a model that must be tested with respect to a given dataset. These two features have served to boost research into the metaheuristics field trying to obtain algorithms needing a small number of fitness function evaluations and so that scale better. Almost all the families of metaheuristics algorithms, from classical to those more recently appeared have been applied to different DM-based problems. Just as an example of this great variety of applied techniques, we can consider the following applications classified by the type of DM task:• Data preprocessing.

show abstract

A new hybrid methodology for cooperative-coevolutionary optimization of radial basis function networks

Cited by 23 publications

References 37 publications

KEEL: a software tool to assess evolutionary algorithms for data mining problems

KEEL: a software tool to assess evolutionary algorithms for data mining problems

CO2RBFN: an evolutionary cooperative–competitive RBFN design algorithm for classification problems

Evolutionary and metaheuristics based data mining

Contact Info

Product

Resources

About