Danillo Roberto Pereira

Robust automated cardiac arrhythmia detection in ECG beat signals

Nowadays, millions of people are affected by heart diseases worldwide, whereas a considerable amount of them could be aided through an electrocardiogram (ECG) trace analysis, which involves the study of arrhythmia impacts on electrocardiogram patterns. In this work, we carried out the task of automatic arrhythmia detection in ECG patterns by means of supervised machine learning techniques, being the main contribution of this paper to introduce the optimum-path forest (OPF) classifier to this context. We compared six distance metrics, six feature extraction algorithms and three classifiers in two variations of the same dataset, being the performance of the techniques compared in terms of effectiveness and efficiency. Although OPF revealed a higher skill on generalizing data, the support vector machines (SVM)-based classifier presented the highest accuracy. However, OPF shown to be more efficient than SVM in terms of the computational time for both training and test phases.

Machine Learning Algorithms for Automatic Classification of Marmoset Vocalizations.

Automatic classification of vocalization type could potentially become a useful tool for acoustic the monitoring of captive colonies of highly vocal primates. However, for classification to be useful in practice, a reliable algorithm that can be successfully trained on small datasets is necessary. In this work, we consider seven different classification algorithms with the goal of finding a robust classifier that can be successfully trained on small datasets. We found good classification performance (accuracy > 0.83 and F1-score > 0.84) using the Optimum Path Forest classifier. Dataset and algorithms are made publicly available.

On the Harmony Search Using Quaternions

Euclidean-based search spaces have been extensively studied to drive optimization techniques to the search for better solutions. However, in high dimensional spaces, non-convex functions might become too tricky to be optimized, thus requiring different representations aiming at smoother fitness landscapes. In this paper, we present a variant of the Harmony Search algorithm based on quaternions, which extend complex numbers and have been shown to be suitable to handle optimization problems in high dimensional spaces. The experimental results in a number of benchmark functions against standard Harmony Search, Improved Harmony Search and Particle Swarm Optimization showed the robustness of the proposed approach. Additionally, we demonstrated the robustness of the proposed approach in the context of fine-tuning parameters in Restricted Boltzmann Machines.

Projections onto convex sets parameter estimation through harmony search and its application for image restoration

Image restoration is a research field that attempts to recover a blurred and noisy image. Although we have one-step algorithms that are often fast for image restoration, iterative formulations allow a better control of the trade-off between the enhancement of high frequencies (image details) and noise amplification. Projections onto convex sets (POCS) is an iterative—and parametric-based approach that employs a priori knowledge about the blurred image to guide the restoration process, with promising results in different application domains. However, a proper choice of its parameters is a high computational burden task, since they are continuous-valued and there are an infinity of possible values to be checked. In this paper, we propose to optimize POCS parameters by means of harmony search-based techniques, since they provide elegant and simple formulations for optimization problems. The proposed approach has been validated in synthetic and real images, being able to select suitable parameters in a reasonable amount of time.

Convolutional Neural Networks Applied for Parkinson’s Disease Identification

Parkinson’s Disease (PD) is a chronic and progressive illness that affects hundreds of thousands of people worldwide. Although it is quite easy to identify someone affected by PD when the illness shows itself (e.g. tremors, slowness of movement and freezing-of-gait), most works have focused on studying the working mechanism of the disease in its very early stages. In such cases, drugs can be administered in order to increase the quality of life of the patients. Since the beginning, it is well-known that PD patients feature the micrography, which is related to muscle rigidity and tremors. As such, most exams to detect Parkinson’s Disease make use of handwritten assessment tools, where the individual is asked to perform some predefined tasks, such as drawing spirals and meanders on a template paper. Later, an expert analyses the drawings in order to classify the progressive of the disease. In this work, we are interested into aiding physicians in such task by means of machine learning techniques, which can learn proper information from digitized versions of the exams, and them recommending a probability of a given individual being affected by PD depending on its handwritten skills. Particularly, we are interested in deep learning techniques (i.e. Convolutional Neural Networks) due to their ability into learning features without human interaction. Additionally, we propose to fine-tune hyper-arameters of such techniques by means of meta-heuristic-based techniques, such as Bat Algorithm, Firefly Algorithm and Particle Swarm Optimization.

Quaternion-based deep belief networks fine-tuning

Deep learning techniques have been paramount in the last years, mainly due to their outstanding results in a number of applications. In this paper, we address the issue of fine-tuning parameters of Deep Belief Networks by means of meta-heuristics in which real-valued decision variables are described by quaternions. Such approaches essentially perform optimization in fitness landscapes that are mapped to a different representation based on hypercomplex numbers that may generate smoother surfaces. We therefore can map the optimization process onto a new space representation that is more suitable to learning parameters. Also, we proposed two approaches based on Harmony Search and quaternions that outperform the state-of-the-art results obtained so far in three public datasets for the reconstruction of binary images.

A Hyperheuristic Approach for Unsupervised Land-Cover Classification

Unsupervised land-use/cover classification is of great interest, since it becomes even more difficult to obtain high-quality labeled data. Still considered one of the most used clustering techniques, the well-known k-means plays an important role in the pattern recognition community. Its simple formulation and good results in a number of applications have fostered the development of new variants and methodologies to address the problem of minimizing the distance from each dataset sample to its nearest centroid (mean). In this paper, we present a genetic programming-based hyperheuristic approach to combine different metaheuristic techniques used to enhance k-means effectiveness. The proposed approach is evaluated in four satellite and one radar image showing promising results, while outperforming each individual metaheuristic technique.

Automatic identification of epileptic EEG signals through binary magnetic optimization algorithms

Epilepsy is a class of chronic neurological disorders characterized by transient and unexpected electrical disturbances of the brain. The automated analysis of the electroencephalogram (EEG) signal can be instrumental for the proper diagnosis of this mental condition. This work presents a systematic assessment of the performance of different variants of the binary magnetic optimization algorithm (BMOA), two of which are introduced here, while serving as feature selectors for epileptic EEG signal identification. In this context, the optimum-path forest classifier was adopted as a classification model, whereas different wavelet families were considered for EEG feature extraction. In order to compare the performance of the improved BMOA variants against the traditional one, as well as other metaheuristic techniques, namely particle swarm optimization, binary bat algorithm, and genetic algorithm, we employed a well-known EEG benchmark dataset composed of five classes of EEG signals (two of which comprising normal patients with eyes open or closed, and the remaining comprising ill patients with different levels of epilepsy). Overall, the results evidenced the robustness of the proposed BMOA and its variants.

Pruning optimum-path forest ensembles using metaheuristic optimization for land-cover classification

ABSTRACT Machine learning techniques have been actively pursued in the last years, mainly due to the increasing number of applications that make use of some sort of intelligent mechanism for decision-making processes. In this context, we shall highlight ensemble pruning strategies, which provide heuristics to select from a collection of classifiers the ones that can really improve recognition rates and provide efficiency by reducing the ensemble size prior to combining the model. In this article, we present and validate an ensemble pruning approach for Optimum-Path Forest (OPF) classifiers based on metaheuristic optimization over general-purpose data sets to validate the effectiveness and efficiency of the proposed approach using distinct configurations in real and synthetic benchmark data sets, and thereafter, we apply the proposed approach in remote-sensing images to investigate the behaviour of the OPF classifier using pruning strategies. The image data sets were obtained from CBERS-2B, LANDSAT-5 TM, IKONOS-2 MS, and GEOEYE sensors, covering some areas of Brazil. The well-known Indian Pines data set was also used. In this work, we evaluate five different optimization algorithms for ensemble pruning, including that Particle Swarm Optimization, Harmony Search, Cuckoo Search, and Firefly Algorithm. In addition, we performed an empirical comparison between Support Vector Machine and OPF using the strategy of ensemble pruning. Experimental results showed the effectiveness and efficiency of ensemble pruning using OPF-based classification, especially concerning ensemble pruning using Harmony Search, which shows to be effective without degrading the performance when applied to large data sets, as well as a good data generalization.

Intrusion Detection System Based On Flows Using Machine Learning Algorithms

The use of technology information and communication by different types of devices generates a large quantity of data packets that contains of confidential and personal information. The traffic of data packet can be summarized in network flow. Due this reason, it is necessary to use computer security tools, such as Intrusion Detection Systems (IDS). This work presents an IDS that can perform the flow- based analysis (netflow). This research conducted an analysis on flows previously collected and properly detected of three different types of attacks. The flows were organized to be processed by machine learning methods. The results obtained by proposed approach were very promising. Also, this work aimed at building a public dataset to be used by researchers worldwide in order to foster IDS-related research.