Abstract The emerging trends in data industry, particularly those related to the repeated process... more Abstract The emerging trends in data industry, particularly those related to the repeated processing of data streams, are pushing the limits of computer systems and processes. Among them, the near real-time clustering of geospatial location data is paradigmatic due to its scale, requirements and potential industrial applications. As a solution to deal with the large and continuous arrival of geospatial data, modern many-core (GPU-based) computer architectures are used for implementing fast and efficient clustering algorithms. This paper proposes GPU implementations of two different clustering algorithms — K -Centres and k -NN mode seeking — and compare them against their corresponding sequential implementations. Publicly available geospatial datasets have been used to exemplify the achieved performances using GPUs. Our main contribution is providing GPU implementations of the clustering algorithms that are feasible for near real-time problems. Results show speed-ups of up to 19 and 135 times, with the largest dataset, for K -Centres and k -NN mode seeking respectively. Important technical details of the sorting algorithms, required by the GPU implementation of k -NN mode seeking, are also highlighted.
Artificial neural networks are not new; they have been around for about 50 years and got some pra... more Artificial neural networks are not new; they have been around for about 50 years and got some practical recognition after the mid-1980s with the introduction of a method (backpropagation) that allowed for the training of multiple-layer neural networks. However, the true birth of deep learning may be traced to the year 2006, when Geoffrey Hinton [GR06] presented an algorithm to efficiently train deep neural networks in an unsupervised way—in other words, data without labels. They were called deep belief networks (DBNs) and consisted of staked restrictive Boltzmann machines (RBMs), with each one placed on the top of another. DBNs differ from previous networks since they are generative models capable of learning the statistical properties of data being presented without any supervision.
Communications in computer and information science, 2015
For financial institutions, the ability to predict or forecast business failures is crucial, as i... more For financial institutions, the ability to predict or forecast business failures is crucial, as incorrect decisions can have direct financial consequences. Credit prediction and credit scoring are the two major research problems in the accounting and finance domain. A variety of pattern recognition techniques including neural networks, decision trees and support vector machines have been applied to predict whether borrowers are in danger of bankruptcy and whether they should be considered a good or bad credit risk.
Recognizing human actions in video sequences, known as Human Action Recognition (HAR), is a chall... more Recognizing human actions in video sequences, known as Human Action Recognition (HAR), is a challenging task in pattern recognition. While Convolutional Neural Networks (ConvNets) have shown remarkable success in image recognition, they are not always directly applicable to HAR, as temporal features are critical for accurate classification. In this paper, we propose a novel dynamic PSO-ConvNet model for learning actions in videos, building on our recent work in image recognition. Our approach leverages a framework where the weight vector of each neural network represents the position of a particle in phase space, and particles share their current weight vectors and gradient estimates of the Loss function. To extend our approach to video, we integrate ConvNets with state-of-the-art temporal methods such as Transformer and Recurrent Neural Networks. Our experimental results on the UCF-101 dataset demonstrate substantial improvements of up to 9% in accuracy, which confirms the effectiv...
This work aims to evaluate the predictive performance of various Machine Learning algorithms when... more This work aims to evaluate the predictive performance of various Machine Learning algorithms when applied to the prediction of material constitutive parameters, particularly the parameters of the Swift hardening law. For this, datasets were generated from the results of the numerical simulations of uniaxial tensile tests. The Machine Learning algorithms considered for this study are: Gaussian Process, Multi-layer Perceptron, Support Vector Regression, Decision Tree and Random Forest. These algorithms were used to train metamodels based on training sets considering different numbers of materials and input parameters, which were then used to predict the hardening law parameters. The Gaussian Process algorithm achieved the overall best predictive performances. The results obtained show the potential of Machine Learning algorithms for application on the identification of material constitutive parameters.
Drug design is an important area of study for pharmaceutical businesses. However, low efficacy, o... more Drug design is an important area of study for pharmaceutical businesses. However, low efficacy, off-target delivery, time consumption, and high cost are challenges and can create barriers that impact this process. Deep Learning models are emerging as a promising solution to perform de novo drug design, i.e., to generate drug-like molecules tailored to specific needs. However, stereochemistry was not explicitly considered in the generated molecules, which is inevitable in targeted-oriented molecules. This paper proposes a framework based on Feedback Generative Adversarial Network (GAN) that includes optimization strategy by incorporating Encoder–Decoder, GAN, and Predictor deep models interconnected with a feedback loop. The Encoder–Decoder converts the string notations of molecules into latent space vectors, effectively creating a new type of molecular representation. At the same time, the GAN can learn and replicate the training data distribution and, therefore, generate new compou...
La deteccion de intrusos en redes de computadoras a partir del enfoque de aprendizaje automatico ... more La deteccion de intrusos en redes de computadoras a partir del enfoque de aprendizaje automatico presenta algunas deficiencias dadas por la propia naturaleza de la aplicacion. La principal viene dada por el modesto despliegue de sistemas de deteccion basados en algoritmos de aprendizaje bajo las restricciones impuestas por los entornos reales. En este articulo se describen y proponen tres variantes de pre procesamiento sobre el conjunto de datos KDD99, incluye seleccion de atributos. Luego la experimentacion se realiza primeramente a partir de evaluar algoritmos representativos en entornos estacionarios sobre las variantes obtenidas a partir de pre procesar KDD99. Por ultimo, dado que el trafico de red es un flujo constante de datos, en el cual pueden existir variaciones de conceptos relacionadas con las tasas de falsos positivos, unido al hecho de que no se encuentran muchas investigaciones que aborden la deteccion de intrusos en entornos de flujos de datos nos conduce a realizar u...
Capsule Networks face a critical problem in computer vision in the sense that the image backgroun... more Capsule Networks face a critical problem in computer vision in the sense that the image background can challenge its performance, although they learn very well on training data. In this work, we propose to improve Capsule Networks' architecture by replacing the Standard Convolution with a Depthwise Separable Convolution. This new design significantly reduces the model's total parameters while increases stability and offers competitive accuracy. In addition, the proposed model on $64\times64$ pixel images outperforms standard models on $32\times32$ and $64\times64$ pixel images. Moreover, we empirically evaluate these models with Deep Learning architectures using state-of-the-art Transfer Learning networks such as Inception V3 and MobileNet V1. The results show that Capsule Networks perform equivalently against Deep Learning models. To the best of our knowledge, we believe that this is the first work on the integration of Depthwise Separable Convolution into Capsule Networks.
The seizure prediction performance of algorithms based in stacked auto-encoders deep-learning tec... more The seizure prediction performance of algorithms based in stacked auto-encoders deep-learning technique has been evaluated. The study is established on long-term electroencephalography (EEG) recordings of 103 patients suffering from drug-resistant epilepsy. The proposed patient-specific methodology consists of feature extraction, classification by machine learning techniques, post-classification alarm generation, and performance evaluation using long-term recordings in a quasi-prospective way. Multiple quantitative features were extracted from EEG recordings. The classifiers were trained to discriminate preictal and non-preictal states. The first part of the feature time series was considered for training, a second part for selection of the “optimal” predictors of each patient, while the remaining data was used for prospective out-of-sample validation. The performance was assessed based on sensitivity and false prediction rate per hour (FPR/h). The prediction performance was statist...
Abstract The emerging trends in data industry, particularly those related to the repeated process... more Abstract The emerging trends in data industry, particularly those related to the repeated processing of data streams, are pushing the limits of computer systems and processes. Among them, the near real-time clustering of geospatial location data is paradigmatic due to its scale, requirements and potential industrial applications. As a solution to deal with the large and continuous arrival of geospatial data, modern many-core (GPU-based) computer architectures are used for implementing fast and efficient clustering algorithms. This paper proposes GPU implementations of two different clustering algorithms — K -Centres and k -NN mode seeking — and compare them against their corresponding sequential implementations. Publicly available geospatial datasets have been used to exemplify the achieved performances using GPUs. Our main contribution is providing GPU implementations of the clustering algorithms that are feasible for near real-time problems. Results show speed-ups of up to 19 and 135 times, with the largest dataset, for K -Centres and k -NN mode seeking respectively. Important technical details of the sorting algorithms, required by the GPU implementation of k -NN mode seeking, are also highlighted.
Artificial neural networks are not new; they have been around for about 50 years and got some pra... more Artificial neural networks are not new; they have been around for about 50 years and got some practical recognition after the mid-1980s with the introduction of a method (backpropagation) that allowed for the training of multiple-layer neural networks. However, the true birth of deep learning may be traced to the year 2006, when Geoffrey Hinton [GR06] presented an algorithm to efficiently train deep neural networks in an unsupervised way—in other words, data without labels. They were called deep belief networks (DBNs) and consisted of staked restrictive Boltzmann machines (RBMs), with each one placed on the top of another. DBNs differ from previous networks since they are generative models capable of learning the statistical properties of data being presented without any supervision.
Communications in computer and information science, 2015
For financial institutions, the ability to predict or forecast business failures is crucial, as i... more For financial institutions, the ability to predict or forecast business failures is crucial, as incorrect decisions can have direct financial consequences. Credit prediction and credit scoring are the two major research problems in the accounting and finance domain. A variety of pattern recognition techniques including neural networks, decision trees and support vector machines have been applied to predict whether borrowers are in danger of bankruptcy and whether they should be considered a good or bad credit risk.
Recognizing human actions in video sequences, known as Human Action Recognition (HAR), is a chall... more Recognizing human actions in video sequences, known as Human Action Recognition (HAR), is a challenging task in pattern recognition. While Convolutional Neural Networks (ConvNets) have shown remarkable success in image recognition, they are not always directly applicable to HAR, as temporal features are critical for accurate classification. In this paper, we propose a novel dynamic PSO-ConvNet model for learning actions in videos, building on our recent work in image recognition. Our approach leverages a framework where the weight vector of each neural network represents the position of a particle in phase space, and particles share their current weight vectors and gradient estimates of the Loss function. To extend our approach to video, we integrate ConvNets with state-of-the-art temporal methods such as Transformer and Recurrent Neural Networks. Our experimental results on the UCF-101 dataset demonstrate substantial improvements of up to 9% in accuracy, which confirms the effectiv...
This work aims to evaluate the predictive performance of various Machine Learning algorithms when... more This work aims to evaluate the predictive performance of various Machine Learning algorithms when applied to the prediction of material constitutive parameters, particularly the parameters of the Swift hardening law. For this, datasets were generated from the results of the numerical simulations of uniaxial tensile tests. The Machine Learning algorithms considered for this study are: Gaussian Process, Multi-layer Perceptron, Support Vector Regression, Decision Tree and Random Forest. These algorithms were used to train metamodels based on training sets considering different numbers of materials and input parameters, which were then used to predict the hardening law parameters. The Gaussian Process algorithm achieved the overall best predictive performances. The results obtained show the potential of Machine Learning algorithms for application on the identification of material constitutive parameters.
Drug design is an important area of study for pharmaceutical businesses. However, low efficacy, o... more Drug design is an important area of study for pharmaceutical businesses. However, low efficacy, off-target delivery, time consumption, and high cost are challenges and can create barriers that impact this process. Deep Learning models are emerging as a promising solution to perform de novo drug design, i.e., to generate drug-like molecules tailored to specific needs. However, stereochemistry was not explicitly considered in the generated molecules, which is inevitable in targeted-oriented molecules. This paper proposes a framework based on Feedback Generative Adversarial Network (GAN) that includes optimization strategy by incorporating Encoder–Decoder, GAN, and Predictor deep models interconnected with a feedback loop. The Encoder–Decoder converts the string notations of molecules into latent space vectors, effectively creating a new type of molecular representation. At the same time, the GAN can learn and replicate the training data distribution and, therefore, generate new compou...
La deteccion de intrusos en redes de computadoras a partir del enfoque de aprendizaje automatico ... more La deteccion de intrusos en redes de computadoras a partir del enfoque de aprendizaje automatico presenta algunas deficiencias dadas por la propia naturaleza de la aplicacion. La principal viene dada por el modesto despliegue de sistemas de deteccion basados en algoritmos de aprendizaje bajo las restricciones impuestas por los entornos reales. En este articulo se describen y proponen tres variantes de pre procesamiento sobre el conjunto de datos KDD99, incluye seleccion de atributos. Luego la experimentacion se realiza primeramente a partir de evaluar algoritmos representativos en entornos estacionarios sobre las variantes obtenidas a partir de pre procesar KDD99. Por ultimo, dado que el trafico de red es un flujo constante de datos, en el cual pueden existir variaciones de conceptos relacionadas con las tasas de falsos positivos, unido al hecho de que no se encuentran muchas investigaciones que aborden la deteccion de intrusos en entornos de flujos de datos nos conduce a realizar u...
Capsule Networks face a critical problem in computer vision in the sense that the image backgroun... more Capsule Networks face a critical problem in computer vision in the sense that the image background can challenge its performance, although they learn very well on training data. In this work, we propose to improve Capsule Networks' architecture by replacing the Standard Convolution with a Depthwise Separable Convolution. This new design significantly reduces the model's total parameters while increases stability and offers competitive accuracy. In addition, the proposed model on $64\times64$ pixel images outperforms standard models on $32\times32$ and $64\times64$ pixel images. Moreover, we empirically evaluate these models with Deep Learning architectures using state-of-the-art Transfer Learning networks such as Inception V3 and MobileNet V1. The results show that Capsule Networks perform equivalently against Deep Learning models. To the best of our knowledge, we believe that this is the first work on the integration of Depthwise Separable Convolution into Capsule Networks.
The seizure prediction performance of algorithms based in stacked auto-encoders deep-learning tec... more The seizure prediction performance of algorithms based in stacked auto-encoders deep-learning technique has been evaluated. The study is established on long-term electroencephalography (EEG) recordings of 103 patients suffering from drug-resistant epilepsy. The proposed patient-specific methodology consists of feature extraction, classification by machine learning techniques, post-classification alarm generation, and performance evaluation using long-term recordings in a quasi-prospective way. Multiple quantitative features were extracted from EEG recordings. The classifiers were trained to discriminate preictal and non-preictal states. The first part of the feature time series was considered for training, a second part for selection of the “optimal” predictors of each patient, while the remaining data was used for prospective out-of-sample validation. The performance was assessed based on sensitivity and false prediction rate per hour (FPR/h). The prediction performance was statist...
Uploads