Coma is a prolonged state of unconsciousness in which a patient exhibits no response to external stimuli. The electroencephalogram (EEG) is a non-invasive exam that measures electrical brain activity through electrodes placed on the scalp, providing real-time insights into neural dynamics. EEG is essential for assessing coma depth, detecting neurological deterioration, and predicting patient outcomes. While deep learning techniques such as convolutional neural networks (CNNs) and long short-term memory (LSTM) networks have been applied to coma prognosis, state-of-the-art architectures like transformers remain largely unexplored in this context. To address this gap, we propose EEGTransformer, a transformer-based architecture designed for the prognostic assessment of comatose patients using EEG signals. EEGTransformer leverages the self-attention mechanism to capture long-range dependencies in EEG sequences while benefiting from multiple attention heads to enhance contextual representation. We evaluate the proposed approach on a real-world dataset comprising dozens of EEG recordings. The results show that EEGTransformer achieves a macro F1 score of 0.88, outperforming state-of-the-art deep learning models, including a CNN-LSTM model and another transformer-based approach. Furthermore, a visual analysis of the learned latent representations reveals a significantly improved class separability compared to existing methods. These findings highlight the potential of transformers for EEG-based biomedical analysis, representing a significant advancement in coma prognosis.

In this paper, we investigate advanced classification techniques for the detection of Autism Spectrum Disorder (ASD) using Attenuated Total Reflectance Fourier Transform Infrared Spectroscopy (ATR-FTIR) of saliva samples. The proposed platform is fast, sustainable, and facilitates non-invasive sample collection. However, the resulting ATR-FTIR spectrum is a high-dimensional sequence that encodes molecular interactions within saliva across an ordered range of wavenumbers, a sequential characteristic that is often neglected in existing studies within the field. While most studies utilizing ATR-FTIR data have employed low-level techniques, which primarily rely on physical attributes such as similarity and distribution of data points, recent research has been conducted using high-level approaches that incorporate topological and structural features in addition to physical characteristics. Despite showing promising results, a critical challenge in these high-level methods lies in the graph construction phase, where current graph construction methods fail to adequately capture the sequential aspects of the input data. This work presents Visibility Graphs Graph (VisG2), a meta-graph construction method designed to map sequential relationships both within a single spectrum and across multiple spectra. Specifically, VisG2 consists of two main components: the representation of each spectrum as a visibility graph and a meta-graph generation procedure based on the similarity between these graphs. Experimental evaluations on a real ASD dataset demonstrate that high-level classification using VisG2 outperforms traditional methods, as well as state-of-the-art ATR-FTIR techniques, including Support Vector Machines and Convolutional Neural Networks. Furthermore, VisG2 surpasses several widely adopted graph construction methods in the literature. Our findings highlight the challenge of effectively capturing sequential relationships in such data and underscore the promising potential of VisG2 for ASD detection and other sequential data problems.

The development of techniques able to forecast the mobile network traffic in a city can feed data driven applications, as Virtual Network Functions (VNF) orchestrators, optimizing the resource allocation and increasing the capacity of mobile networks. Despite the fact that several studies have addressed this problem, many did not consider neither the traffic relationship among city regions nor the information retrieved from public transport stations, which may provide useful information to better anticipate the network traffic. In this paper, we propose a new deep learning based architecture to forecast the network traffic using representation learning and recurrent neural networks. The framework, named Mobile Traffic Predictor Enhanced by Neighboring and Transportation Data (MTP-NT), has two major components: the first one is responsible of learning from the time series of the region to be predicted, with the second one learning from the time series of both neighboring regions and public transportation stations. Several experiments were conducted over a dataset from the city of Milan, as well as comparisons against widely adopted and state-of-the-art techniques. The results shown in this paper demonstrate that the usage of public transport information contributes to improve the forecasts in central areas of the city, as well as in regions with aperiodic demands, such as tourist regions.

The dynamics of forest plantations have been widely studied with computational simulation applications. Cellular automata (CA) is a technique capable of modelling future states based on a set of transition rules. However, this construction is not simple, often requiring technical knowledge of the process through years of scientific research. Machine learning techniques can be applied in this context, facilitating the construction of these simulators. This work presents a simulation model based on probabilistic cellular automata capable of estimating the evolution of wood production throughout the management period. Unlike other works in the literature, the construction of the CA transition rule is based exclusively on historical data from a Tachi-branco plantation, a managed forest species. Linear and logistic regression models are applied to learn and represent the local transition rules of the automaton and simulate its evolution. The proposed CA-based approach was able to predict the future behavior of plantations in the monitored areas with errors around 4%, confirming the potential of using machine learning in discovering transition rules for precise models.

The design of peptides capable of inhibiting the SARS-CoV-2 viral infection has been considered one of the potential strategies to reduce the transmission of SARS-CoV-2. However, one critical issue in peptide design is the large search space, which makes it impracticable to evaluate all possibilities. Furthermore, most related analyses adopt in silico molecular docking to select potential peptides, which is a time-consuming technique and highly dependent on the molecular structure of already known peptides and the target protein. Aiming to assist the evaluation, discovery and selection of peptides for docking calculation, we developed SAGAPEP, a Surrogate-Assisted Genetic Algorithm framework capable of finding peptides with potential to block the SARS-Co V-2 Spike protein. The surrogate model is used for fast and high-fidelity evaluation of the interaction energy between a peptide and the Spike protein, while the genetic algorithm seeks to discover and select high-potential peptides inspired by principles of genetics and natural selection. Experiments were conducted using a data set composed of several potential peptides obtained through molecular docking by bio-informatics specialists. Our experimental results demonstrate that SAGAPEP achieved low error predictions from its surrogate component trained over that data set, and was able to discover and select peptides with higher binding energy than all present in the data set. Moreover, the noteworthy results of SAGAPEP suggest it may also have the potential to provide promising results for other peptide design problems.

This work investigates high-level classification techniques based on properties and measures of complex networks for the salivary detection of oral cancer from Attenuated Total Reflectance by Fourier Transform Infrared Spectroscopy (ATR-FTIR). Saliva biomarkers are alternative to surrogate other invasive samples in the early detection and monitoring of systemic diseases. Saliva also allows convenient and easy collection. ATR-FTIR is a sustainable, rapid and non-invasive platform able to contribute to the detection of several diseases. Traditional machine learning techniques have already been considered in the analysis of salivary ATR-FTIR data. However, such techniques are able to perform only low-level classification of the spectra data by considering physical features such as similarity, distance or distribution. On the other hand, high-level techniques are able to consider the semantic meaning of the input data by analyzing their structural and topological properties. In this paper, we investigate the hypothesis that the high-level classification of the ATR-FTIR spectra obtained via learning systems based on complex networks measures can achieve better predictive performance in comparison with low-level classification techniques widely adopted in the literature, such as linear discriminant analysis and support vector machines (SVM). Experiments conducted on real-world data confirmed our hypothesis. Our high-level classification techniques achieved 71% of accuracy and 81% of sensitivity in the detection of oral cancer after considering properties and structural patterns captured by the Clustering Coefficient network measure. Moreover, such results also outperformed those obtained by state-of-the-art classifiers like convolutional neural networks.

Coffee pricing is strongly influenced by fruit quality. High-quality coffee fruits are achieved when developed under adequate nutritional, climatic, and health conditions but also by finding the appropriate harvest time. This paper aimed to present a methodological approach based on Convolutional Neural Networks (CNN) to identify the ripening process of coffee fruits in order to support farmers in the harvest decision. The study was conducted using images of coffee plants from farms located in Brazilian cities which were further labelled by several experts regarding its maturation stage as well as the appropriateness to harvest or not. Such images were later used in the learning and evaluation of state-of-the-art CNNs architecture models. The computer simulations results were satisfactory, with the model surpassing 92% accuracy, thereby achieving values higher than those of some current models. This approach can significantly improve harvest management by increasing the precision of fruit ripening classification systems and the predictability of crop evolution, thereby increasing both the added value of the product and reducing costs through quality management.

High-level classification are supervised learning techniques able to consider topological and structural features of the input data. Several high-level techniques have been proposed in the last years with different strategies, such as the pattern conformation technique which represents the input data as a network and perform classification by analyzing the variation of complex network measures. Such techniques have contributed in several tasks, but their contribution to the classification of sequential patterns has not been investigated yet. In this paper, we propose a high-level technique based on the complex network measures assortativity and average shortest path length to the analysis of electroencephalogram (EEG) data. To be specific, we consider two formulation of the problem of prognosis of patients in coma: binary and multi-class. The problem is very difficult and challenging as the data contains patients from different etiologies. Experimental results with nine other techniques including state-of-the-art ones like convolutional neural networks revealed that our high-level approach has the potential to improve (statistically) the predictive performance of those techniques, especially when considering the results with the assortativity measure for both binary and multi-class formulations. Moreover, this study paves a way in the adoption of complex network measures besides the extraction of features from EEG records, but also for the classification itself.

This work investigates a classification technique based on centrality properties of complex networks. Different from traditional classifiers which consider only the physical features of the data (e.g., similarity or distribution), the technique under study also considers structural and topological features of the networked data. In previous studies the technique takes into account the individual importance of each input data in the classification of new instances by adopting the well-known pagerank measure, while other relevant centrality measures were not even considered. In this paper we cover such a lacuna by analyzing a total of five relevant centrality measures from the literature, namely: pagerank, betweenness, closeness, degree and shortest path length. Such measures had their bias evaluated over several real-world data sets in terms of predictive capability and robustness. The results showed that pagerank and degree often achieved the best results and also outperformed statistically all other measures in terms of predictive robustness. In a few words, these findings may support both the understanding and appropriate selection of complex network measures for machine learning tasks.

Coma can be caused by different health conditions. Sometimes, patients are admitted to intensive care unit (ICU) without the cause of the coma being known. Knowing the coma etiology of a patient is very important for prognosis and treatment. Classification of electroencephalogram (EEG) signals by deep learning is proposed to help predict the coma etiology of ICU patients. EEG is a cheap noninvasive technique that can be used for the diagnostics and evaluation of neurological diseases. The objective is to classify coma etiology into one of four categories: Traumatic Brain Injury (TBI), Metabolic Coma, Stroke, and Other. A deep learning model based on convolutional neural networks (CNNs) is proposed to classify the EEG signals, using information from two different sources: i) intermediate layers of CNN + long-short term memory network (LSTM); ii) additional features from patients and statistical measures extracted from EEG signals. Outputs of the LSTM and additional features are inserted as additional inputs to the first dense layer of the CNN. The proposed model was compared to six other approaches, some of which incorporated additional features from patients or statistical measures from EEG signals, while others did not. Experimental results show that inserting patient information, like age and genre, as input to the first dense layer improve the predictive performance of the classification model. Moreover, this work suggests new possibilities to assist physicians in the detection of the coma etiology, especially those in small and far health units.

Complex networks have become an increasingly relevant research topic in machine learning, with many learning systems in the literature successfully exploring complex network properties and measures. In data classification, the use of complex networks allows the detection of structural and topological patterns related, for example, to the formation pattern of the input data. Some measures of complex networks have already been used in this sense. However, a systematic study capable of characterizing such measures in the context of data classification is lacking in the literature. In this work, we evaluate comparatively the predictive performance of some measures. Specifically, eight complex network measures were selected from the literature, namely: assortativity, average local clustering coefficient, average degree, betweenness, average shortest path length, closeness, global clustering coefficient and eigenvector centrality. For our analyses, both artificial and real-world data sets were considered. The results show that measures such as average shortest path and assortativity, besides presenting high predictive capability, are also more robust to the variation of the network structure. In summary, this research paves a way to support other related works in selecting more appropriate complex network measures for data classification.

Multi-label learning (MLL) addresses the problem of learning from data items which can be associated with multiple labels simultaneously. As MLL techniques are usually derived of single-label ones, they also share common drawbacks. For example, most MLL techniques perform a low-level classification, i.e., they consider only the physical features of the input data (e.g., distance, distribution, etc) in the classification process, having troubles to detect semantic relationships among the data items, like the formation pattern for example. Recent studies have shown that learning systems based on complex networks have the ability to consider not only the physical features of the data, but also structural and topological features extracted from the network connection patterns, which is known as high-level classification. In this paper, we investigate a MLL framework which combines both low-level and high-level techniques in order to improve the predictive performance of existing MLL techniques. Experiments conducted on artificial and real-world data sets highlighted the salient features of the MLL framework and also attested its good predictive performance in comparison with widely used MLL techniques, indicating that our framework may considerable improve their predictive performance.

Although the packing of executable binaries can be adopted with legitimate intent such as intellectual property protection and size reduction, malware developers utilize those tools to obfuscate their code and thus increase the complexity of static analysis. In order to recognize packed executables, the BinStat application was proposed. It is based on two major steps: the feature extraction, which involves the calculation of statistics and information theory properties from a given binary; and the classification, which adopts a decision tree learned from input features of packed and unpacked binaries previously known in order to classify new executables. The results obtained proved the effectiveness of the tool, but the choice of using only one classifier is arguably a weakness that we chose to improve on the present study. For that end, we rebuilt the training and test datasets and selected the following six classifiers to our analyses: classification and regression trees, random forest, k-nearest neighbors, naive Bayes, neural network and support vector machines. Our results show that the original decision tree algorithm adopted in BinStat (C5.0) is not the best choice for the proposed problem. Indeed, random forest, k-nearest neighbors and support vector machines achieved the best predictive performances.

Multi-label learning aims to solve problems in which data items can have multiple class labels assigned simultaneously, e.g., text categorization, image annotation, medical diagnosis, etc. However, as most of multi-label techniques are derived from the single-label ones, existing techniques perform the multi-label classification only based on the physical features of the data (e.g., distance, similarity or distribution), ignoring the semantic meaning of the data, such as the formation pattern. Inspired by recent advances in the use of complex networks for single-label learning, this exploratory work aims to investigate a multi-label solution able to combine existing multi-label classifiers with a high-level classifier based on complex networks measures, aiming to present a new concept of multi-label classification that, besides the physical attributes, also analyzes the topological structure of the data. Experimental results considering both artificial and real-world data sets emphasize respectively the salient features of our technique in comparison to the traditional ones and its potential to improve the predictive performance of those techniques, especially in data sets characterized by higher cardinality and density of labels, which often denote more difficult scenarios to multi-label learning.

Complex networks provide a powerful tool for data representation due to its ability to describe the interplay between topological, functional, and dynamical properties of the input data. A fundamental process in network-based (graph-based) data analysis techniques is the network construction from original data usually in vector form. Here, a natural question is: How to construct an "optimal" network regarding a given processing goal? This paper investigates structural optimization in the context of network-based data classification tasks. To be specific, we propose a particle swarm optimization framework which is responsible for building a network from vector-based data set while optimizing a quality function driven by the classification accuracy. The classification process considers both topological and physical features of the training and test data and employing PageRank measure for classification according to the importance concept of a test instance to each class. Results on artificial and real-world problems reveal that data network generated using structural optimization provides better results in general than those generated by classical network formation methods. Moreover, this investigation suggests that other kinds of network-based machine learning and data mining tasks, such as dimensionality reduction and data clustering, can benefit from the proposed structural optimization method.

In this article, we address the problem of modeling the actions of a human player in order to learn his strategies from his past game logs in Zoom Texas Hold'em poker variant. Although Texas Hold'em is a very popular game, Zoom is yet a very recent format of game in which, instead of playing in a specific table against a specific set of opponents, a player is placed in a large pool of players in which their opponents change every hand. Pros and cons of Zoom include respectively bigger effective time playing (and possibly getting money) and scarcity of data to get reads from the opponents. To deal with this problem, our model consists of a simple and generic set of features designed to fulfill each one of four proposed categories (hand quality, position insights, aggressiveness and current situation) in order to be able to capture a wide range of player strategies in each stage of the game. As a consequence of our modeling, we generate five data sets which were further evaluated by machine learning techniques. The results show that much of the player strategies were effectively learned, especially by non-linear techniques. Moreover, our data sets are available online as a test-bed for machine learning research in poker games.

Supervised classification techniques are known to exploit physical information of the analysed data, such as similarity, distribution and other low level features. Despite the relevance of such features, recent works have showed that a higher variety of patterns can be detected by combining low level and high level features. In this paper, it is proposed a supervised classification technique which applies limiting probabilities of the random walk theory over underlying networks constructed from input labeled data. The appealing feature of the proposed approach is that the adjacency matrix which carries both physical and structural information about the data. Structural information are given by features extracted from network connections. The class of a given unlabeled sample is estimated by a heuristic called ease of access, which is measured by the random walk process over the adjacency matrix. Such approach makes the technique quite general as one can put distinct data measures of interest in the connection matrix of the underlying data network to guide the random walker. Specifically, we show examples of combining low and high level features in the proposed classification scheme. Simulation results using artificial and real data sets suggest that the proposed technique is not only competitive with current and established classification techniques, but it also can reveal intrinsic structural patterns formed by the input data.

Data classification is a common task, which can be performed by both computers and human beings. However, a fundamental difference between them can be observed: computer-based classification considers only physical features (e.g., similarity, distance, or distribution) of input data; by contrast, brain-based classification takes into account not only physical features, but also the organizational structure of data. In this paper, we figure out the data organizational structure for classification using complex networks constructed from training data. Specifically, an unlabeled instance is classified by the importance concept characterized by Google's PageRank measure of the underlying data networks. Before a test data instance is classified, a network is constructed from vector-based data set and the test instance is inserted into the network in a proper manner. To this end, we also propose a measure, called spatio-structural differential efficiency, to combine the physical and topological features of the input data. Such a method allows for the classification technique to capture a variety of data patterns using the unique importance measure. Extensive experiments demonstrate that the proposed technique has promising predictive performance on the detection of heart abnormalities.

Graph-based dimensionality reduction has attracted a lot of attention in recent years. Such methods aim to exploit the graph representation in order to catch some structural information hidden in data. They usually consist of two steps: graph construction and projection. Although graph construction is crucial to the performance, most research work in the literature has focused on the development of heuristics and models to the projection step, and only very recently, attention was paid to network construction. In this work, graph construction is considered in the context of supervised dimensionality reduction. To be specific, using a nature-inspired optimization framework, this work investigates if an optimized graph is able to provide better projections than well-known general-purpose methods. The proposed method is compared with widely used graph construction methods on a range of real-world image classification problems. Results show that the optimization framework has achieved considerable dimensionality reduction rates as well as good predictive performance.

While most part of the complex network models are described in function of some growth mechanism, the optimization of a goal or certain characteristics can be desirable for some problems. This paper investigates structural optimization of networks in the highlevel classification context, where the classification produced by a traditional classifier is combined with the classification provided by complex network measures. Using the recently proposed social learning particle swarm optimization (SL-PSO), a bio-inspired optimization framework, which is responsible to build up the network and adjust the parameters of the hybrid model while conducting the optimization of a quality function, is proposed. Experiments on two real-world problems, the Handwritten Digits Recognition and the Semantic Role Labeling (SRL), were performed. In both problems, the optimization framework is able to improve the classification given by a state-of-the-art algorithm to SRL. Furthermore, the optimization framework proposed here can be extended to other machine learning tasks.

In machine learning, dimensionality reduction aims at reducing the dimension of the input data in order to achieve a small set of features that keeps the most important original relationships among data samples. In this paper, we investigate the usage of a non-parametric network formation algorithm into a graph embedding framework to perform supervised dimensionality reduction. Specifically, our technique maps data into networks and constructs two network adjacency matrices which convey information about intra-class components and inter-class penalty connections. Both matrices are inserted into an optimization framework in order to achieve a projection vector that is used to project high-dimension data samples into a low-dimensional space. One advantage of the technique is that no parameter is required, that is, there is no need to select a model for the input data. Computer simulations on real-world data sets have been performed to compare the proposed technique to some classical network formation methods such as k-NN and e-radius, and to well-known dimensionality reduction algorithms such as PCA and LDA. Statistical tests have shown that our approach outperforms those algorithms.