Navegando por Autor "Torres, Luiz Carlos Bambirra"

Agora exibindo 1 - 15 de 15

An advanced pruning method in the architecture of extreme learning machines using L1-regularization and bootstrapping.
(2020) Souza, Paulo Vitor de Campos; Torres, Luiz Carlos Bambirra; Silva, Gustavo Rodrigues Lacerda; Braga, Antônio de Pádua; Lughofer, Edwin
Extreme learning machines (ELMs) are efficient for classification, regression, and time series prediction, as well as being a clear solution to backpropagation structures to determine values in intermediate layers of the learning model. One of the problems that an ELM may face is due to a large number of neurons in the hidden layer, making the expert model a specific data set. With a large number of neurons in the hidden layer, overfitting is more likely and thus unnecessary information can deterioriate the performance of the neural network. To solve this problem, a pruning method is proposed, called Pruning ELM Using Bootstrapped Lasso BR-ELM, which is based on regularization and resampling techniques, to select the most representative neurons for the model response. This method is based on an ensembled variant of Lasso (achieved through bootstrap replications) and aims to shrink the output weight parameters of the neurons to 0 as many and as much as possible. According to a subset of candidate regressors having significant coefficient values (greater than 0), it is possible to select the best neurons in the hidden layer of the ELM. Finally, pattern classification tests and benchmark regression tests of complex real-world problems are performed by comparing the proposed approach to other pruning models for ELMs. It can be seen that statistically BR-ELM can outperform several related state-of-the-art methods in terms of classification accuracies and model errors (while performing equally to Pruning-ELM P-ELM), and this with a significantly reduced number of finally selected neurons.
Combined weightless neural network FPGA architecture for deforestation surveillance and visual navigation of UAVs.
(2020) Torres, Vitor Angelo Maria Ferreira; Jaimes, Brayan Rene Acevedo; Ribeiro, Eduardo da Silva; Braga, Mateus Taulois; Shiguemori, Elcio Hideit; Velho, Haroldo Fraga de Campos; Torres, Luiz Carlos Bambirra; Braga, Antônio de Pádua
This work presents a combined weightless neural network architecture for deforestation surveillance and visual navigation of Unmanned Aerial Vehicles (UAVs). Binary images, which are required for position estimation and UAV navigation, are provided by the deforestation surveillance circuit. Learned models are evaluated in a real UAV flight over a green countryside area, while deforestation surveillance is assessed with an Amazon forest benchmarking image data. Small utilization percentage of Field Programmable Gate Arrays (FPGAs) allows for a higher degree of parallelization and block processing of larger regions of input images.
Combined weightless neural network FPGA architecture for deforestation surveillance and visual navigation of UAVs.
(2020) Torres, Vitor Angelo Maria Ferreira; Jaimes, Brayan Rene Acevedo; Ribeiro, Eduardo S.; Braga, Mateus Taulois; Shiguemori, Elcio Hideiti; Velho, Haroldo Fraga de Campos; Torres, Luiz Carlos Bambirra; Braga, Antônio de Pádua
This work presents a combined weightless neural network architecture for deforestation surveillance and visual navigation of Unmanned Aerial Vehicles (UAVs). Binary images, which are required for position estimation and UAV navigation, are provided by the deforestation surveillance circuit. Learned models are evaluated in a real UAV flight over a green countryside area, while deforestation surveillance is assessed with an Amazon forest benchmarking image data. Small utilization percentage of Field Programmable Gate Arrays (FPGAs) allows for a higher degree of parallelization and block processing of larger regions of input images.
Data density-based clustering for regularized fuzzy neural networks based on nullneurons and robust activation function.
(2019) Souza, Paulo Vitor de Campos; Torres, Luiz Carlos Bambirra; Guimarães, Augusto Júnio; Araújo, Vanessa Souza; Araújo, Vinicius Jonathan Silva; Rezende, Thiago Silva
This paper proposes the use of fuzzification functions based on clustering of data based on their density to perform the granularization of the input space. The neurons formed in this layer are built through the density centers obtained with the input data of the model. In the second layer, the nullneurons aggregate the generated neurons in the first layer and allow the creation of if/then fuzzy rules. Even in the second layer, a regularization function is activated to determine the essential nullneurons. The concepts of extreme learning machine generate the weights used in the third layer, but with a regularizing factor. Finally, in the third layer, represented by an artificial neural network, it has a single neuron that the activation function uses robust functions to carry out the model. To verify the new training approach for fuzzy neural networks, we performed real and synthetic database tests for the pattern classification, which led to the conclusion that the data density-based approach the use of regularization factors in the second model layer and neurons with more robust activation functions allowed better results compared to other classifiers that use the concepts of extreme learning machine.
Embedded real-time feature extraction for electrode inversion detection in telemedicine electrocardiograms.
(2020) Torres, Vitor Angelo Maria Ferreira; Silva, D. A.C.; Torres, Luiz Carlos Bambirra; Braga, Mateus Taulois; Cardoso, Mathues B. R.; Lino, Vinicius Terra; Torres, Frank Sill; Braga, Antônio de Pádua
Early detection of technical errors in medical examinations, especially in remote locations, is of utmost importance in order to avoid invalid measurements that would require costly and time consuming repeti- tions. This paper proposes a highly efficient method for the identification of an erroneous inversion of the measuring electrodes during a multichannel electrocardiogram. Therefore, a widely applied approach for heart beat detection is modified and approximated feature extraction techniques are employed. In con- trast to existing works, the improved heart beat identification requires no removal of baseline wandering and no amplitude related thresholds. Furthermore, a piecewise linear approximation of the baseline and basic calculations are sufficient for extracting the cardiac axis, which allows the construction of a clas- sifier capable of quickly detecting electrode reversals. Our implementation indicates that the proposed method has minimal hardware costs and is able to operate in real-time on a simple micro-controller.
Enhancing performance of Gabriel graph-based classifiers by a hardware co-processor for embedded system applications.
(2020) Arias Garcia, Janier; Mafra, Augusto Amaral; Gade, Liliane dos Reis; Coelho, Frederico Gualberto Ferreira; Castro, Cristiano Leite de; Torres, Luiz Carlos Bambirra; Braga, Antônio de Pádua
It is well known that there is an increasing interest in edge computing to reduce the distance between cloud and end devices, especially for Machine Learning (ML) methods. However, when related to latency-sensitive applications, little work can be found in ML literature on suitable embedded systems implementations. This paper presents new ways to implement the decision rule of a large margin classifier based on Gabriel graphs as well as an efficient implementation of this on an embedded system. The proposed approach uses the nearest neighbor method as the decision rule, and the implementation starts from an RTL pipeline architecture developed for binary large margin classifiers and proposes the integration in a hardware/software co-design. Results showed that the proposed approach was statistically similar to the classifier and had a speedup factor of up to 8x compared to the classifier executed in software, with performance suitable for ML latency-sensitive applications.
Extreme wavelet fast learning machine for evaluation of the default profle on financial transactions.
(2020) Souza, Paulo Vitor de Campos; Torres, Luiz Carlos Bambirra
Extreme learning machines enable multilayered neural networks to perform activities to facilitate the process and business dynamics. It acts in pattern classifcation, linear regression problems, and time series prediction. The fnancial area needs efcient models that can perform businesses in a short time. Credit card fraud and debits occur regularly, and efective decision making can avoid signifcant obstacles for both clients and fnancial companies. This paper proposes a training model for multilayer networks where the weights of the training algorithm are defned by the nature and characteristics of the dataset using the concepts of the wavelet transform. The traditional algorithm of weights’ defnition of the output layer is changed to a regularized method that acts more quickly in the description of the weights of the output layer. Finally, several activation functions are applied to the model to verify its efciency in several scenarios. This model was subjected to an extensive dataset and comparing to diferent machine learning approaches. Its answers were satisfactory in a short-time execution, proving that the Extreme Learning Machine works effciently to identify possible profles of defaulters in payments in the fnancial relationships involving a credit card.
A fuzzy data reduction cluster method based on boundary information for large datasets.
(2019) Silva, Gustavo Rodrigues Lacerda; Cirino Neto, Paulo; Torres, Luiz Carlos Bambirra; Braga, Antônio de Pádua
The fuzzy c-means algorithm (FCM) is aimed at computing the membership degree of each data point to its corresponding cluster center. This computation needs to calculate the distance matrix between the cluster center and the data point. The main bottleneck of the FCM algorithm is the computing of the membership matrix for all data points. This work presents a new clustering method, the bdrFCM (boundary data reduction fuzzy c-means). Our algorithm is based on the original FCM proposal, adapted to detect and remove the boundary regions of clusters. Our implementation efforts are directed in two aspects: processing large datasets in less time and reducing the data volume, maintaining the quality of the clusters. A significant volume of real data application ([106 records) was used, and we identified that bdrFCM implementation has good scalability to handle datasets with millions of data points.
Large margin gaussian mixture classifier with a Gabriel graph geometric representation of data set structure.
(2020) Torres, Luiz Carlos Bambirra; Castro, Cristiano Leite de; Coelho, Frederico Gualberto Ferreira; Braga, Antônio de Pádua
This brief presents a geometrical approach for obtaining large margin classifiers. The method aims at exploring the geometrical properties of the data set from the structure of a Gabriel graph, which represents pattern relations according to a given distance metric, such as the Euclidean distance. Once the graph is generated, geometrical support vectors (SVs) (analogous to support vector machines (SVMs) SVs) are obtained in order to yield the final large margin solution from a Gaussian mixture model. Experiments with 20 data sets have shown that the solutions obtained with the proposed method are statistically equivalent to those obtained with SVMs. However, the present method does not require optimization and can also be extended to large data sets using the cascade SVM concept.
Multi-objective neural network model selection with a graph-based large margin approach.
(2022) Torres, Luiz Carlos Bambirra; Castro, Cristiano Leite de; Rocha, Honovan Paz; Almeida, Gustavo Matheus de; Braga, Antônio de Pádua
This work presents a new decision-making strategy for multi-objective learning problem of artificial neural networks (ANN). The proposed decision-maker searches for the solution that minimizes a margin-based validation error amongst Pareto set solutions. The proposal is based on a geometric approximation to find the large margin (distance) of separation among the classes. Several benchmarks commonly available in the literature were used for testing. The obtained results showed that the proposal is more efficient in controlling the generalization capacity of neural models than other learning machines. It yields smooth (noise robustness) and well-fitted models straightforwardly, i.e., without the necessity of parameter set definition in advance or validation data use, as often required by learning machines.
Neural networks regularization with graph-based local resampling.
(2021) Assis, Alex Damiany; Torres, Luiz Carlos Bambirra; Araújo, Lourenço Ribeiro Grossi; Hanriot, Vítor Mourão; Braga, Antônio de Pádua
This paper presents the concept of Graph-based Local Resampling of perceptron-like neural networks with random projections (RN-ELM) which aims at regularization of the yielded model. The addition of synthetic noise to the learning set finds some similarity with data augmentation approaches that are currently adopted in many deep learning strategies. With the graph-based approach, however, it is possible to direct resample in the margin region instead of exhaustively cover the whole input space. The goal is to train neural networks with added noise in the margin region, located by structural information extracted from a planar graph. The so-called structural vectors, which are the training set vertices near the class boundary, are obtained from the structural information using Gabriel Graph. Synthetic samples are added to the learning set around the geometric vectors, improving generalization performance. A mathematical formulation that shows that the addition of synthetic samples has the same effect as the Tikhonov regularization is presented. Friedman and pos-hoc Nemenyi tests indicate that outcomes from the proposed method are statistically equivalent to the ones obtained by objective-function regularization, implying that both methods yield smoother solutions, reducing the effects of overfitting.
Pulsar detection for wavelets soda and regularized fuzzy neural networks based on andneuron and robust activation function.
(2019) Souza, Paulo Vitor de Campos; Torres, Luiz Carlos Bambirra; Guimarães, Augusto Júnio; Araújo, Vanessa Souza
The use of intelligent models may be slow because of the number of samples involved in the problem. The identification of pulsars (stars that emit Earth-catchable signals) involves collecting thousands of signals by professionals of astronomy and their identification may be hampered by the nature of the problem, which requires many dimensions and samples to be analyzed. This paper proposes the use of hybrid models based on concepts of regularized fuzzy neural networks that use the representativeness of input data to define the groupings that make up the neurons of the initial layers of the model. The andneurons are used to aggregate the neurons of the first layer and can create fuzzy rules. The training uses fast extreme learning machine concepts to generate the weights of neurons that use robust activation functions to perform pattern classification. To solve large-scale problems involving the nature of pulsar detection problems, the model proposes a fast and highly accurate approach to address complex issues. In the execution of the tests with the proposed model, experiments were conducted explanation in two databases of pulsars, and the results prove the viability of the fast and interpretable approach in identifying such involved stars.
Utilizando aprendizado por representação para a classificação de laços sociais da IoT
(2022) Pereira Júnior, Jamisson Jader Moraes; Figueiredo, Thiago Silva; Lopes, Ramon; Torres, Luiz Carlos Bambirra; Santos, Bruno P
A Internet of Things (IoT) tem sido marcada pelas interações entre dispositivos que cooperam para realizar atividades. A partir deste ambiente cibernético e conectado, um possível paradigma derivado é o Social IoT (SIoT), onde múltiplos tipos de relacionamentos e confiabilidade podem ser estabelecidos entre dispositivos. Neste cenário, abordamos as questões de como modelar lac¸os sociais em IoT e na proposição de modelos para, automaticamente, classificar e predizer relações em SIoT. Este artigo propõe a utilização de aprendizado por representação para classificar diferentes tipos de lac¸os sociais em SIoT. Para isso, utiliza-se como estratégias para classificação Graph Neural Networks (GNN) ou Algoritmos Tradicionais de Classificação (ATC). Em nossos experimentos, GNN é rápido na etapa de treinamento e apresenta métricas F1-{macro, micro} de 0.61 e 0.88, respectivamente. Ao usar ATC, o treinamento ´e 121× at´e 11.235× mais lento que GNN, ao passo que as métricas F1-score alcançam 0.86 e 0.95, respetivamente.
Wearables and detection of falls : a comparison of machine learning methods and sensors positioning.
(2022) Pinto, Arthur Bernardo Assumpção; Assis, Gilda Aparecida de; Torres, Luiz Carlos Bambirra; Beltrame, Thomas; Domingues, Diana Maria Gallicchio
Wearable sensors have many applications to provide assistance for older adults. We aimed to identify the best combination of machine learning algorithms and body regions to attach one wearable for real-time falls detection from a public dataset where volunteers performed daily activities and simulated falls. Accuracy and comfort of the combination of wearables and algorithms were assessed. Raw data from the accelerometer and gyroscope were used for both training and testing stages. We evaluated the confusion matrix between all wear- ables at each of the different body regions (Ankle, Right Pocket, Belt, Neck, and Wrist) for the following machine learning algorithms: Multilayer Perceptron (MLP), Random Forest, XGBoost, and Long Short Term Memory (LSTM) deep neural network. The accuracy was compared by ANOVA two-way repeated measures statistical test. This work has two main technical contributions. First, our results demonstrated the highest accuracy in identifying falls when the sensors were positioned on the neck or ankle. Second, when the machine learning algorithms to detect fall was compared, LSTM deep neural network and Random Forest showed statistically higher accuracy than MLP and XGBoost. Besides, a comfort analysis based on the literature concluded that neck and wrist are the most comfortable regions to wear wearables.
Width optimization of RBF kernels for binary classification of support vector machines : a density estimation-based approach.
(2019) Menezes, Murilo V. F.; Torres, Luiz Carlos Bambirra; Braga, Antônio de Pádua
Kernels are often used for modelling non-linear data, developing a main role in models like the SVM. The optimization of its parameters to better fit each dataset is a frequently faced challenge: A bad choice of kernel parameters often implies a poor model. This problem is usually worked out using exhaustive search approaches, such as cross-validation. These methods, however, do not take into account existent information on data arrangement. This paper proposes an alternative approach, based on density estimation. By making use of density estimation methods to analyze the dataset structure, it is proposed a function over the kernel parameters. This function can be used to choose the parameters that best suit the data.