-
[show abstract]
[hide abstract]
ABSTRACT: This work proposes a methodology for sleep stage classification based on two main approaches: the combination of features extracted from electroencephalogram (EEG) signal by different extraction methods, and the use of stacked sequential learning to incorporate predicted information from nearby sleep stages in the final classifier. The feature extraction methods used in this work include three representative ways of extracting information from EEG signals: Hjorth features, wavelet transformation and symbolic representation. Feature selection was then used to evaluate the relevance of individual features from this set of methods. Stacked sequential learning uses a second-layer classifier to improve the classification by using previous and posterior first-layer predicted stages as additional features providing information to the model. Results show that both approaches enhance the sleep stage classification accuracy rate, thus leading to a closer approximation to the experts' opinion.
Electronic version of an article published as [International Journal of Neural Systems, Volume 23, Issue 3, Year 2013, Pages 1350012 -20 pages] [DOI: 10.1142/S0129065713500123] © [copyright World Scientific Publishing Company] [http://www.worldscientific.com/worldscinet/ijns]
International Journal of Neural Systems 06/2013; 23(3):1350012. · 4.28 Impact Factor
-
[show abstract]
[hide abstract]
ABSTRACT: This paper presents a first approach to try to determine if a newborn will be macrosomic before the labor, using a set of data taken from the mother. The problem of determining if a newborn is going to be macrosomic is important in order to plan cesarean section and other problems during the labor. The proposed model to classify the weight is a neural network whose design is based recent algorithms that will allow the networks to focus on a concrete class. Before proceeding with the design methodology to obtain the models, a previous step of variable selection is performed in order to indentify the risk factors and to avoid the curse of dimensionality. Another study is made regarding the missing values in the database since the data were not complete for all the patients. The results will show how useful the addition of the missing values into the original data set can be in order to identify new risk factors.
Intelligent Systems Design and Applications, 2009. ISDA '09. Ninth International Conference on; 01/2010
-
[show abstract]
[hide abstract]
ABSTRACT: The kernel weighted k-nearest neighbours (KWKNN) algorithm is an efficient kernel regression method that achieves competitive results with lower computational complexity than least-squares support vector machines and Gaussian processes. This paper presents the parallel implementation on a cluster platform of the sequential KWKNN implemented in Matlab. This implies both the parallelization of the k nearest-neighbour search and the evaluation of the cross-validation error on a large distributed data set. The results demonstrate the good performances of the implementation.
High Performance Computing & Simulation, 2009. HPCS '09. International Conference on; 07/2009
-
[show abstract]
[hide abstract]
ABSTRACT: Traditionally, the autoregressive moving average (ARMA) model has been one of the most widely used linear models in time series prediction. Recent research activities in forecasting with artificial neural networks (ANNs) suggest that ANNs can be a promising alternative to the traditional ARMA structure. These linear models and ANNs are often compared with mixed conclusions in terms of the superiority in forecasting performance. In this paper we propose a hybridization of intelligent techniques such as ANNs, fuzzy systems and evolutionary algorithms, so that the final hybrid ARIMA–ANN model could outperform the prediction accuracy of those models when used separately. More specifically, we propose the use of fuzzy rules to elicit the order of the ARMA or ARIMA model, without the intervention of a human expert, and the use of a hybrid ARIMA–ANN model that combines the advantages of the easy-to-use and relatively easy-to-tune ARIMA models, and the computational power of ANNs.
Fuzzy Sets and Systems. 01/2008;
-
[show abstract]
[hide abstract]
ABSTRACT: Radial basis function neural networks (RBFNNs) are well known because, among other applications, they present a good performance when approximating functions although their design still remains as a difficult task. The function approximation problem arises in the construction of a control system to optimize the process of the mineral reduction. In order to regulate the temperature of the ovens and other parameters, a module to predict the final concentration of mineral that will be obtained from the source materials is necessary. In a previous work, this problem was successfully solved by designing an RBFNN using a MultiObjective genetic algorithm (MOGA). However, the more samples are obtained from the system, the more difficult it becomes to design the RBFNN due to the high dimensionality of the problem. Therefore, a new algorithm that addresses the dimensionality reduction has been developed, allowing to obtain more accurate RBFNNs, deciding which input parameters must be considered. Another important element incorporated in the algorithm is the concept of fuzzy dominance, the algorithm, when performing the sorting of the population dividing it in subsets of non-dominated individuals, uses a fuzzy criteria to decide if an individual dominates another. As the experimental results will show, the new version of the algorithm generates RBFNNs with smaller approximation errors and less complexity due to the reduction in the number of input variables and neurons.
Fuzzy Systems Conference, 2007. FUZZ-IEEE 2007. IEEE International; 08/2007
-
[show abstract]
[hide abstract]
ABSTRACT: There are many papers in the literature that deal with the problem of the design of a fuzzy system from a set of given training
examples. Those who get the best approximation accuracy are based on TSK fuzzy rules, which have the problem of not being
as interpretable as Mamdany-type Fuzzy Systems. A question now is posed: How can the interpretability of the generated fuzzy
rule-table base be increased? A possible response is to try to reduce the rule-base size by generalizing fuzzy-rules consequents
which are symbolic functions instead of fixed scalar values or polynomials, and apply symbolic regressions technics in fuzzy
system generation. A first approximation to this idea is presented in this paper for 1-D and 2D functions.
06/2005: pages 137-148;
-
[show abstract]
[hide abstract]
ABSTRACT: In this paper we design an on-line controller which is able to modify and adapt the rule base of the system with just only qualitative knowledge about the plant to be controlled. Since flying a helicopter is an extremely difficult task, the fuzzy logic controller was necessarily quite complex. In fact, the control tasks were distributed over four individual control units, each of which had its own rules and associated membership functions. Because the fuzzy logic controller was large, and because the rules implemented in the individual control units were not necessarily those a human pilot would use, an efficient technique for writing the rules was required. A genetic algorithm was used to discover rules that provided for effective control of the helicopter. Our study is focused on the module responsible for controlling the helicopter's altitude. For the simulations performed by the adaptive controller, we modify, in a dynamic way, the value of the mass of the helicopter. This would correspond, in real life, to an increase or decrease, for example, in the number of passengers, discharge of water in a fire, etc. On the basis of the nominal value of the helicopter's mass, various simulations are performed to modify the latter parameter within a 15% range. Faced with such a situation, the values of the consequences of the rules responsible for controlling the helicopter's altitude must vary, as otherwise it would not be possible to maintain a zero difference between the desired altitude and that measured by the sensors. Finally, due to the speed requirement, the controller is implemented in FPGA
Advanced Motion Control, 2006. 9th IEEE International Workshop on; 02/2001
-
[show abstract]
[hide abstract]
ABSTRACT: This paper presents a new methodology to achieve real time self tuning and self learning in fuzzy controllers, with application to motion control of a trailer for reaching an aiming point and obstacle-avoidance. The advantage of this approach is that it only requires qualitative information about the plant to be controlled, in terms of the monotony presented by the output with respect to the control signal and the delay of the plant. Also, starting with a non-optimum controller, the system is able to self-adapt its behaviour in order to reduce the error. Thus, it is capable of controlling highly non-linear systems, in a pseudo-optimum way, even when these are time variable, for example, the dynamic of the robot or trailer change (i.e.: different mass, different environments, different dynamics of the system). Control is achieved by means of two auxiliary systems: the first one is responsible for adapting the consequences of the main controller to minimize the error arising at the plant output, while the second auxiliary system compiles real input/output data obtained from the plant. The methodology has been successfully applied to a real robot with different dynamics and in different environments, showing its ability to tune its +behaviour
Advanced Motion Control, 2006. 9th IEEE International Workshop on; 02/2001
-
[show abstract]
[hide abstract]
ABSTRACT: Radial basis function neural networks (RBFNNs) have been applied to solve problems of classification, function approximation and time series prediction. In the design of an RBFNN it is necessary to set the values for the positions of the centers and the radii for each RBF. In the literature it is usually performed an initialization step to set the positions of the centers and, once they are placed, the radii are calculated using a heuristic. In this paper, a new algorithm to set the value of those two parameters is presented. This new algorithm uses a supervised learning in such a way that the position of the centers will be constrained by the output of the function to be approximated. Since each center represents a neuron that is activated by the input vectors, the radii are initialized using the center's positions and their activation grades. In this way, the calculation of the radii is also influenced by the output of the target function, not like in other heuristics where only the positions of the centers or the input vectors are considered. As the experiments show, the new algorithm outperforms other algorithms previously used for this problem.
Neural Networks, 2006. IJCNN '06. International Joint Conference on;
-
[show abstract]
[hide abstract]
ABSTRACT: The problem of selecting the patterns to be learned by any model is usually not considered by the time of designing the concrete model but as a preprocessing step. Information theory provides a robust theoretical framework for performing input variable selection thanks to the concept of mutual information. Recently the computation of the mutual information for regression tasks has been proposed so this paper presents a new application of the concept of mutual information not to select the variables but to decide which prototypes should belong to the training data set in regression problems. The proposed methodology consists in deciding if a prototype should belong to or not to the training set using as criteria the estimation of the mutual information between the variables. The novelty of the approach is to focus in prototype selection for regression problems instead of classification as the majority of the literature deals only with the last one. Other element that distinguishes this work from others is that it is not proposed as an outlier detector but as an algorithm that determines the best subset of input vectors by the time of building a model to approximate it. As the experiment section shows, this new method is able to identify a high percentage of the real data set when it is applied to highly distorted data sets.
Neurocomputing.
-
[show abstract]
[hide abstract]
ABSTRACT: The hierarchical network proposed (Multi-RBFNN), is composed of complete Radial Basis Function Neural Networks (RBFNNs) that are in charge of a reduced set of input variables with the property of which every Sub-RBFNN can take charge of a set of input variables and not of all. For the optimization of the whole net, we propose a new method to select the more important input variables, which is capable of deciding which of the chosen variables go alone or together to a Sub-RBFNN to build the hierarchic structure Multi-RBFNN, thus reducing the dimension of the input variable space for each RBFNN. We also provide an algorithm which automatically finds the most suitable topology of the proposed hierarchical structure and selects the more important input variables for it.
