Combining Meta-learning and Search Techniques to SVM Parameter Selection

Gomes, Taciana A. F.; Prudêncio, Ricardo B. C.; Soares, Carlos; Rossi, André Luis Debiaso; Carvalho, André C. P. L. F. de

doi:10.1109/sbrn.2010.22

Cited by 10 publications

(8 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, the cost to obtain a response from a meta-learning model is usually lower than the use of search algorithms. There is also an intermediate approach in which meta-learning is applied to provide initial parameter configurations to be refined by a search algorithm (as performed in [11]). In this case, the search algorithm would start from a promising region in the search space (provided by meta-learning) and could find satisfactory solutions faster.…”

Section: Resultsmentioning

confidence: 99%

“…In this study, a set of 42 meta-examples was generated from 42 regression datasets. In [11], meta-learning was used to optimize two SVM parameters: the parameter γ of the RBF kernel and the regularization constant C. A set of 40 meta-examples was produced from different regression problems. An instancebased learner was used in the meta-level to recommend the SVM parameters among a set of 399 candidate combinations of γ and C.…”

Section: Meta-learningmentioning

confidence: 99%

“…The meta-knowledge acquired from the set of meta-examples is used to predict which is the best configuration of parameters based on the characteristics of the learning problems at hand. The meta-learning techniques were adopted to optimize parameters of algorithms, more specifically, the parameters of Support Vector Machines (SVM) [9] [10] [11]. The good performance of meta-learning in these works motivates the use of meta-learning in other models of ANNs.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Meta-learning to optimize the number of hidden nodes of MLP networks trained by Extreme Learning Machine algorithm

Lucas¹,

Prudêncio²,

Ludermir³

et al. 2016

Anais Do 10. Congresso Brasileiro De Inteligência Computacional

View full text Add to dashboard Cite

Section: Resultsmentioning

confidence: 99%

Section: Meta-learningmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Meta-learning to optimize the number of hidden nodes of MLP networks trained by Extreme Learning Machine algorithm

Lucas¹,

Prudêncio²,

Ludermir³

et al. 2016

Anais Do 10. Congresso Brasileiro De Inteligência Computacional

View full text Add to dashboard Cite

“…For instance, in [5], the authors presented a meta-learning approach to predict the accuracy of two algorithms, multi-layer perceptron with backpropagation and with Levenberg-Marquardt. On the other hand, in [6], the authors use a hybrid approach with meta-learning and search algorithms in order to automatically adjust the parameters of a support vector machine (SVM). In [7], the authors used meta-learning to build a model to predict a ranking of performance among the main learning algorithms used in gene expression tasks.…”

Section: Introductionmentioning

confidence: 99%

An Empirical Analysis of Meta-learning for the Automatic Choice of Architecture and Components in Ensemble Systems

Nascimento

Canuto

Coelho

2014

2014 Brazilian Conference on Intelligent Systems

View full text Add to dashboard Cite

Studies with ensemble systems have gained attention recently and, most of them, propose new methods for the design (generation) of different components in these systems. In parallel, new contributions of meta-learning have been presented as an efficient alternative to automatic recommendation of algorithms. In this paper, we apply meta-learning in the process of recommendation of important parameters of ensemble systems, which are: architecture and individual classifiers. The main goal is to provide an efficient way to design ensemble systems. In order to validate the proposed approach, an empirical investigation is conducted, recommending three possible types of ensemble architectures (Bagging, Boosting and Multi-Boosting) and five possible types of learning algorithms to compose the ensemble systems (individual classifiers or components). An initial analysis of the results confirms that meta-learning can be a promising tool to be used in the automatic choice of important parameters in ensemble systems.

show abstract

“…In a recent work [19], ML and search techniques were combined for SVM parameter selection. In this work, ML was adopted to suggest a number of solutions (configurations of parameters) which are adopted as the initial population of the search technique.…”

Section: Introductionmentioning

confidence: 99%

Combining a multi-objective optimization approach with meta-learning for SVM parameter selection

Miranda

Prudêncio

Carvalho

et al. 2012

2012 IEEE International Conference on Systems, Man, and Cybernetics (SMC)

View full text Add to dashboard Cite

Support Vector Machine (SVM) is a supervised technique, which achieves good performance on different learning problems. However, adjustments on its model are essentials to the SVM work well. Optimization techniques have been used to automatize this process finding suitable configurations of parameters which attends some learning problems. This work utilizes Particle Swarm Optimization (PSO) applied to the SVM parameter selection problem. As the learning systems are essentially a multi-objective problem, a multi-objective PSO (MOPSO) was used to maximize the success rate and minimize the number of support vectors of the model. Nevertheless, we propose the combination of Meta-Learning (ML) with a modified MOPSO which uses the crowding distance mechanism (MOPSO-CDR). In this combination, solutions provided by ML are possibly located in good regions in the search space. Hence, using a reduced number of successful candidates, the search process would converge faster and be less expensive. In our work, we implemented a prototype in which MOPSO-CDR was used to select the values of two SVM parameters for classification problems. In the performed experiments, the proposed solution (MOPSO-CDR using ML) was compared to the MOPSO-CDR with random initialization, obtaining pareto fronts with higher quality on a set of 40 classification problems.

show abstract

Combining Meta-learning and Search Techniques to SVM Parameter Selection

Cited by 10 publications

References 14 publications

Meta-learning to optimize the number of hidden nodes of MLP networks trained by Extreme Learning Machine algorithm

Meta-learning to optimize the number of hidden nodes of MLP networks trained by Extreme Learning Machine algorithm

An Empirical Analysis of Meta-learning for the Automatic Choice of Architecture and Components in Ensemble Systems

Combining a multi-objective optimization approach with meta-learning for SVM parameter selection

Contact Info

Product

Resources

About