Análise de ferramentas computacionais utilizadas em modelos numéricos para representação de processos atmosféricos, oceânicos e de superfície

A transformação digital é uma realidade em todas as áreas da atividade humana e vem sendo acelerada pela recente pandemia da Covid-19. O uso de Inteligência Artificial, Data Analytics, Realidade Aumentada, Robots, dentre outras tecnologias viabiliza a automação de processos praticamente em todos os segmentos da economia: agricultura, manufatura e serviços. Entender os fenômenos climáticos e seus impactos no ambiente sempre foi um desafio para o homem que, desde a antiguidade e até os dias atuais, se depara com situações não previstas, que trazem impacto em seu dia a dia, colocando, muitas vezes, a própria vida em risco. Na atualidade, além de buscar entender, o que se busca é prever o clima e seus impactos ambientais, de modo a mitigar os riscos, viabilizar a continuidade do desenvolvimento, preservando a natureza (fauna, flora, recursos naturais etc). O uso de ferramentas computacionais para modelagem numérica que representem este fenômenos é de suma importância, endereçando a necessidade de previsões cada vez mais rápidas e confiáveis. Este projeto analisa ferramentas disponíveis no mercado, que utilizam três modelos numéricos distintos, descrevendo as funcionalidades existentes e a assertividade na previsão de fenômenos climáticos em Santa Catarina

Orthogonal decomposition of the sum-symmetry model for square contingency tables with ordinal categories: Use of the exponential sum-symmetry model

Summary In the existing decomposition theorem, the sum-symmetry model holds if and only if both the exponential sum-symmetry and global symmetry models hold. However, this decomposition theorem does not satisfy the asymptotic equivalence for the test statistic. To address the aforementioned gap, this study establishes a decomposition theorem in which the sum-symmetry model holds if and only if both the exponential sum-symmetry and weighted global-sum-symmetry models hold. The proposed decomposition theorem satisfies the asymptotic equivalence for the test statistic. We demonstrate the advantages of the proposed decomposition theorem by applying it to datasets comprising real data and artificial data.

Comparison of GANs for Covid-19 X-ray classification

Image classification has been applied to several real problems. However, getting labeled data is a costly task, since it demands time, resources and experts. Furthermore, some domains like disease detection suffer from unbalanced classes. These scenarios are challenging and degrade the performance of machine learning algorithms. In these cases, we can use Data Augmentation (DA) approaches to increase the number of labeled examples in a dataset. The objective of this work is to analyze the use of Generative Adversarial Networks (GANs) as DA, which are capable of synthesizing artificial data from the original data, under an adversarial process of two neural networks. The GANs are applied in the classification of unbalanced Covid-19 radiological images. Increasing the number of images led to better accuracy for all the GANs tested, especially in the multi-label dataset, mitigating the bias for unbalanced classes.

Diagnosis of Diabetic Retinopathy through Retinal Fundus Images and 3D Convolutional Neural Networks with Limited Number of Samples

Diabetic retinopathy (DR) is a worldwide problem associated with the human retina. It leads to minor and major blindness and is more prevalent among adults. Automated screening saves time of medical care specialists. In this work, we have used different deep learning (DL) based 3D convolutional neural network (3D-CNN) architectures for binary and multiclass (5 classes) classification of DR. We have considered mild, moderate, no, proliferate, and severe DR categories. We have deployed two artificial data augmentation/enhancement methods: random weak Gaussian blurring and random shift along with their combination to accomplish these tasks in the spatial domain. In the binary classification case, we have found the performance of 3D-CNN architecture trained by deploying combined augmentation methods to be the best, while in the multiclass case, the performance of model trained without augmentation is the best. It is observed that the DL algorithms working with large volumes of data may achieve better performances as compared to the methods working with small volumes of data.

A Safe Zone SMOTE Oversampling Algorithm Used in Earthquake Prediction Based on Extreme Imbalanced Precursor Data

Earthquake prediction based on extreme imbalanced precursor data is a challenging task for standard algorithms. Since even if an area is in an earthquake-prone zone, the proportion of days with earthquakes per year is still a minority. The general method is to generate more artificial data for the minority class that is the earthquake occurrence data. But the most popular oversampling methods generate synthetic samples along line segments that join minority class instances, which is not suitable for earthquake precursor data. In this paper, we propose a Safe Zone Synthetic Minority Oversampling Technique (SZ-SMOTE) oversampling method as an enhancement of the SMOTE data generation mechanism. SZ-SMOTE generates synthetic samples with a concentration mechanism in the hyper-sphere area around each selected minority instances. The performance of SZ-SMOTE is compared against no oversampling, SMOTE and its popular modifications adaptive synthetic sampling (ADASYN) and borderline SMOTE (B-SMOTE) on six different classifiers. The experiment results show that the quality of earthquake prediction using SZ-SMOTE as oversampling algorithm significantly outperforms that of using the other oversampling algorithms.

Flexible Covariance Matrix Decomposition Method for Data Augmentation and Its Application to Brainwave Signals

The acquisition of a large-volume brainwave database is challenging because of the stressful experiments that are required; however, data synthesis techniques can be used to address this limitation. Covariance matrix decomposition (CMD), a widely used data synthesis approach, generates artificial data using the correlation between features and random noise. However, previous CMD methods constrain the stochastic characteristics of artificial datasets because the random noise used follows a standard distribution. Therefore, this study has improved the performance of CMD by releasing such constraints. Specifically, a generalized normal distribution (GND) was used as it can alter the kurtosis and skewness of the random noise, affecting the distribution of the artificial data. For the validation of GND performance, a motor imagery brainwave classification was conducted on the artificial dataset generated by GND. The GND-based data synthesis increased the classification accuracy obtained with the original data by approximately 8%.

Anomaly detection in genomic catalogues using unsupervised multi-view autoencoders

Background Accurate identification of Transcriptional Regulator binding locations is essential for analysis of genomic regions, including Cis Regulatory Elements. The customary NGS approaches, predominantly ChIP-Seq, can be obscured by data anomalies and biases which are difficult to detect without supervision. Results Here, we develop a method to leverage the usual combinations between many experimental series to mark such atypical peaks. We use deep learning to perform a lossy compression of the genomic regions’ representations with multiview convolutions. Using artificial data, we show that our method correctly identifies groups of correlating series and evaluates CRE according to group completeness. It is then applied to the ReMap database’s large volume of curated ChIP-seq data. We show that peaks lacking known biological correlators are singled out and less confirmed in real data. We propose normalization approaches useful in interpreting black-box models. Conclusion Our approach detects peaks that are less corroborated than average. It can be extended to other similar problems, and can be interpreted to identify correlation groups. It is implemented in an open-source tool called atyPeak.

Inferring Functional Epigenetic Modules by Integrative Analysis of Multiple Heterogeneous Networks

Gene expression and methylation are critical biological processes for cells, and how to integrate these heterogeneous data has been extensively investigated, which is the foundation for revealing the underlying patterns of cancers. The vast majority of the current algorithms fuse gene methylation and expression into a network, failing to fully explore the relations and heterogeneity of them. To resolve these problems, in this study we define the epigenetic modules as a gene set whose members are co-methylated and co-expressed. To address the heterogeneity of data, we construct gene co-expression and co-methylation networks, respectively. In this case, the epigenetic module is characterized as a common module in multiple networks. Then, a non-negative matrix factorization-based algorithm that jointly clusters the co-expression and co-methylation networks is proposed for discovering the epigenetic modules (called Ep-jNMF). Ep-jNMF is more accurate than the baselines on the artificial data. Moreover, Ep-jNMF identifies more biologically meaningful modules. And the modules can predict the subtypes of cancers. These results indicate that Ep-jNMF is efficient for the integration of expression and methylation data.

Magnetic Field Sensors’ Calibration: Algorithms’ Overview and Comparison

The calibration of three-axis magnetic field sensors is reviewed. Seven representative algorithms for in-situ calibration of magnetic field sensors without requiring any special piece of equipment are reviewed. The algorithms are presented in a user friendly, directly applicable step-by-step form, and are compared in terms of accuracy, computational efficiency and robustness using both real sensors’ data and artificial data with known sensor’s measurement distortion.

Which Index Should I Use? A Comparison of Indices for Enclosure Use Studies

Enclosure use assessments have gained popularity as one of the tools for animal welfare assessments and Post Occupancy Evaluations. There are now a plethora of studies and enclosure use indices available in published literature, and identification of the most appropriate index for each research question is often challenging. The benefits and limitations of four different enclosure use indices; Original and Modified Spread of Participation Index, Entropy, and Electivity Index were compared. Three artificial data sets were developed to represent the challenges commonly found in animal exhibits, and these indices were applied to these contrived enclosure settings. Three of the indices (Original SPI, Modified SPI, and Entropy) were used to assess a single measure of enclosure use variability. When zones within an exhibit were comparable in size, all three indices performed similarly. However, with less equal zone sizes, Modified SPI outperformed Original SPI and Entropy, suggesting that the Modified formula was more useful for assessing overall enclosure use variability under such conditions. Electivity Index assessed the use of individual zones, rather than the variability of use across the entire exhibit, and therefore could not be compared directly to the other three indices. This index is therefore most valuable for assessing individual resources, especially after exhibit modifications.