Entity

Time filter

Source Type

Maracanaú, Brazil

Rocha Neto A.R.,Federal Institute of Ceara | Barreto G.A.,Federal University of Ceara
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2011

We introduce a novel heuristic based on the Kohonen's SOM, called Opposite Maps, for building reduced-set SVM classifiers. When applied to the standard SVM (trained with the SMO algorithm) and to the LS-SVM method, the corresponding reduced-set classifiers achieve equivalent (or superior) performances than standard full-set SVMs. © 2011 Springer-Verlag.


Da Rocha Neto A.R.,Federal Institute of Ceara | Barreto G.A.,Federal University of Ceara
Advances in Intelligent Systems and Computing | Year: 2013

This paper introduces a new approach to building hard margin classifiers based on Opposite Maps (OM). OM is a Self-Organizing Map-based method used for obtaining reduced-set classifiers in the sense of soft margin. As originally proposed, Opposite Maps was used for reducing the training data set and obtaining soft margin reduced-set SVM and LSSVM classifiers. In our new proposal we use Opposite Maps in order to obtain a set of patterns in the overlapping area between positive and negative classes and, a posteriori, to remove them from the default training data set. This process can transform a non-linear problem into a linear one in which a hard-margin classifier like Huller SVM can be applied. This approach assure to get resulting classifiers from a training process without needing to set up the cost parameter C that controls the trade off between allowing training errors and margin maximization. Besides that, but differently from soft-margin classifiers, these obtained classifiers leave the patterns at wrong side of the hyperplane out of the set of support vectors and, therefore, reduced-set hard-margin classifiers come out with few support vectors. © 2013 Springer-Verlag.


Da Rocha Neto A.R.,Federal Institute of Ceara | Barreto G.A.,Federal University of Ceara
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2012

Opposite Maps (OM) is a method that can be used to induce sparse SVM-based and LS-SVM-based classifiers. The main idea behind the OM method is to train two Self-Organizing Maps (SOM), one for each class, and , in a binary classification context and then, for the patterns of one class, say , to find the closest prototypes among those belonging to the SOM trained with patterns of the other class, say . The subset of patterns mapped to the selected prototypes in both SOMs form the reduced set to be used for training SVM and LSSVM classifiers. In this paper, an iterative method based on the OM, called Fast Opposite Maps, is introduced with the aim of accelerating OM training time. Comprehensive computer simulations using synthetic and real-world datasets reveal that the proposed approach achieves similar results to the original OM, at a much faster pace. © 2012 Springer-Verlag.


Silva R.V.,Federal University of Ceara | Freitas A.A.A.,Federal University of Ceara | Castro M.R.,Federal University of Ceara | Antunes F.L.M.,Rural University | Sa E.M.,Federal Institute of Ceara
Conference Proceedings - IEEE Applied Power Electronics Conference and Exposition - APEC | Year: 2016

This paper proposes a non-isolated qZS converter to feed a frequency inverter applied on a standalone photovoltaic tricycle. The converter uses the clipping voltage of the power switch to be added to the output voltage, which allows reducing the duty cycle values and increase the gain of the converter. Some converter topologies, based on high gain coupled inductors and voltage multiplier cells cause large current ripple in the input and voltage spike at the power switch. As a consequence the lifespan of the components and the converter efficiency are reduced. The proposed converter showed a low current ripple in the input and the power switch was submitted to lower voltage. High gain qZS dc-dc converters operation principle and analysis are shown, and verified by simulation and experimental results as well as a losses analysis. © 2016 IEEE.


Paiva Mesquita D.P.,Federal University of Ceara | Gomes J.P.P.,Federal University of Ceara | Souza A.H.,Federal Institute of Ceara
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2015

Minimal Learning Machine (MLM) is a recently proposed supervised learning algorithm with simple implementation and few hyper parameters. Learning MLM model consists on building a linear mapping between input and output distance matrices. In this work, the standard MLM is modified to deal with missing data. For that, the expected squared distance approach is used to compute the input space distance matrix. The proposed approach showed promising results when compared to standard strategies that deal with missing data. © Springer International Publishing Switzerland 2015.

Discover hidden collaborations