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.

Abstract Character Coreference Resolution in Movie Screenplays (MovieCoref) is a newly emerging task for understanding complex movie plots and character relationships. This task poses greater challenges than traditional coreference resolution, due to the intricate narrative structures and character interactions unique to screenplays. In light of these challenges, we introduce a novel approach: a Structure-aware Triaffine Parser (STPar) for the MovieCoref task. STPar combines discourse and syntactic structures in the feature encoding process, enabling comprehensive analysis of ternary relationships and complex interactions. During the pairing process, STPar utilizes a triaffine scorer to consider high-order relations between candidate mention pairs, thus enhancing its ability to capture detailed narrative structures. In addition, STPar incorporates multi-task learning, encompassing singleton and span detection tasks, to further improve coreference resolution performance. Our evaluations on the MovieCoref dataset demonstrate that STPar significantly outperforms the best baseline by 7.4%, 21.5%, 7.1%, and 10.2% in F1 scores of B3, CEAFe, LEA, and CoNLL. Further analysis highlights the benefits of integrating structural discourse and syntactic information as well as the combined approaches of triaffine and multi-task learning.1

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.

The field of Graph Neural Networks (GNNs) has garnered significant attention for leveraging Neural Networks to analyze graph data. However, the inherent limitations of GNNs, arising from their aggregation function and injectivity constraints, restrict their ability to distinguish various topological neighborhood structures accurately. Consequently, after aggregation, different underlying graph structures may not be distinctly identified, impeding the model’s representation capabilities. In this study, we introduce a novel GNN model that incorporates Self-Organizing Maps (SOM) to enhance the representation of node attributes and characterize the graph’s neighborhood structure. Leveraging SOM, we transform the node’s attribute space into a fixed-size grid, enabling the representation of neighborhood topology by overlaying activated regions, which allows the model to learn association rules based on the presence or absence of neighboring attributes. This process generates an injective graph embedding that uniquely captures different neighborhood structures. We evaluate the model on real-world datasets to assess its’ performance, analyzing how SOM parameters influence its behavior and effectiveness. Through extensive experimentation, we demonstrate that our model achieves competitive and robust results with well-established GNN models. Our approach showcases the potential of using SOM to overcome limitations in current GNN architectures, providing a more effective and nuanced representation of graph data.

Few-shot named entity recognition (NER) exploits limited annotated instances to identify named mentions. Effectively transferring the internal or external resources thus becomes the key to few-shot NER. While the existing prompt tuning methods have shown remarkable few-shot performances, they still fail to make full use of knowledge. In this work, we investigate the integration of rich knowledge to prompt tuning for stronger few-shot NER. We propose incorporating the deep prompt tuning framework with threefold knowledge (namelyTKDP), including the internal 1)context knowledgeand the external 2)label knowledge& 3)sememe knowledge. TKDP encodes the three feature sources and incorporates them into soft prompt embeddings, which are further injected into an existing pre-trained language model to facilitate predictions. On five benchmark datasets, the performance of our knowledge-enriched model was boosted by at most 11.53% F1 over the raw deep prompt method, and it significantly outperforms 9 strong-performing baseline systems in 5-/10-/20-shot settings, showing great potential in few-shot NER. Our TKDP framework can be broadly adapted to other few-shot tasks without much effort.

Graph Neural Networks (GNNs) have become a rapidly growing field, due to their ability to capture the relationship among data, instead of only learning from the attribute of the data. The core of any GNN is the graph embedding generation by message passing mechanisms. In this work we propose a new message passing technique based on the Particle Competition and Cooperation (PCC) model, originally developed for community detection in graphs. The proposed framework performs a transductive learning in the network and passes the learned information to the nodes, prior to the inductive learning performed by traditional GNN schemes. The new GNN presents attractive features which overcomes the over-smoothing problem of traditional GNNs and shows promising results in terms of classification accuracy, computational cost and learning with very small quantity of labeled data.

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.

So far, discontinuous named entity recognition (NER) has received increasing research attention and many related methods have surged such as hypergraph-based methods, span-based methods, and sequence-to-sequence (Seq2Seq) methods, etc. However, these methods more or less suffer from some problems such as decoding ambiguity and efficiency, which limit their performance. Recently, grid-tagging methods, which benefit from the flexible design of tagging systems and model architectures, have shown superiority to adapt for various information extraction tasks. In this paper, we follow the line of such methods and propose a competitive grid-tagging model for discontinuous NER. We call our model TOE because we incorporate two kinds of Tag-Oriented Enhancement mechanisms into a state-of-the-art (SOTA) grid-tagging model that casts the NER problem into word-word relationship prediction. First, we design a Tag Representation Embedding Module (TREM) to force our model to consider not only word-word relationships but also word-tag and tag-tag relationships. Concretely, we construct tag representations and embed them into TREM, so that TREM can treat tag and word representations as queries/keys/values and utilize self-attention to model their relationships. On the other hand, motivated by the Next-Neighboring-Word (NNW) and Tail-Head-Word (THW) tags in the SOTA model, we add two new symmetric tags, namely Previous-Neighboring-Word (PNW) and Head-Tail-Word (HTW), to model more fine-grained word-word relationships and alleviate error propagation from tag prediction. In the experiments of three benchmark datasets, namely CADEC, ShARe13 and ShARe14, our TOE model pushes the SOTA results by about 0.83%, 0.05% and 0.66% in F1, demonstrating its effectiveness.

Event extraction (EE) is an essential task of information extraction, which aims to extract structured event information from unstructured text. Most prior work focuses on extracting flat events while neglecting overlapped or nested ones. A few models for overlapped and nested EE includes several successive stages to extract event triggers and arguments,which suffer from error propagation. Therefore, we design a simple yet effective tagging scheme and model to formulate EE as word-word relation recognition, called OneEE. The relations between trigger or argument words are simultaneously recognized in one stage with parallel grid tagging, thus yielding a very fast event extraction speed. The model is equipped with an adaptive event fusion module to generate event-aware representations and a distance-aware predictor to integrate relative distance information for word-word relation recognition, which are empirically demonstrated to be effective mechanisms. Experiments on 3 overlapped and nested EE benchmarks, namely FewFC, Genia11, and Genia13, show that OneEE achieves the state-of-the-art (SOTA) results. Moreover, the inference speed of OneEE is faster than those of baselines in the same condition, and can be further substantially improved since it supports parallel inference.

Link prediction (LP) in networks aims at determining future interactions among elements; it is a critical machine-learning tool in different domains, ranging from genomics to social networks to marketing, especially in e-commerce recommender systems. Although many LP techniques have been developed in the prior art, most of them consider only static structures of the underlying networks, rarely incorporating the network's information flow. Exploiting the impact of dynamic streams, such as information diffusion, is still an open research topic for LP. Information diffusion allows nodes to receive information beyond their social circles, which, in turn, can influence the creation of new links. In this work, we analyze the LP effects through two diffusion approaches, susceptible-infected-recovered and independent cascade. As a result, we propose the progressive-diffusion (PD) method for LP based on nodes' propagation dynamics. The proposed model leverages a stochastic discrete-time rumor model centered on each node's propagation dynamics. It presents low-memory and low-processing footprints and is amenable to parallel and distributed processing implementation. Finally, we also introduce an evaluation metric for LP methods considering both the information diffusion capacity and the LP accuracy. Experimental results on a series of benchmarks attest to the proposed method's effectiveness compared with the prior art in both criteria.

An important task in unsupervised learning is the detection of communities in networks. Although many community detection techniques have been proposed, there are still some challenge problems, such as unbalanced community detection and the low efficiency. In this paper, we propose a community detection technique combining the sequential signal propagation of the Particle Competition model and the parallel propagation inspired by Self-Orgnizing Map (SOM). As a result, the model presents two salient features: 1) It can detect unbalanced communities. 2) It is much more efficient than the original particle competition model due to the introduction of parallel propagation. Still in this work, we analyze functional brain network by identifying the modules (communities) using the proposed technique. Our results show that there is a strong correlation between brain functions and brain regions and a big decrease of intra-strength measure among communities from the Control Network to the Schizophrenia Network, indicating that the functional correlation of brain regions is weakened in the disease network.

In this paper, we introduce a network-based classification model which, instead of mapping each data instance as a node in a network, as usual, it maps each data instance attribute as being a node. This procedure allows the model to preserve more information from the input dataset when building the network, specially for datasets with a larger number of features, and thus to make use of this extra information during the training phase. In addition, we also introduce a technique intended to generate a network with one component per class in the dataset while keeping the threshold parameter, which is responsible for determining the edges among the nodes, at a minimum value. In this way, the network emerging from this process is more sensitive to the insertion of new instances, during the testing phase, in terms of its modularity measure, which allows the classifier to infer the new labels based mainly on this measure. We evaluate the model by applying it to both artificial and real benchmark classification datasets, and have its performance compared to those obtained by other traditional classification models on the same data. The preliminary results are encouraging, with the proposed model being ranked on second place among the 10 classifiers considered, on the selected datasets.

Chat groups are well-known for their capacity to promote viral political and marketing campaigns, spread fake news, and create rallies by hundreds of thousands on the streets. Also, with the increasing public awareness regarding privacy and surveillance, many platforms have started to deploy end-to-end encrypted protocols. In this context, the group's conversations are not accessible in plain text or readable format by third-party organizations or even the platform owner. Then, the main challenge that emerges is related to getting insights from users' activity of those groups, but without accessing the messages. Previous approaches evaluated the user engagement by assessing user's activity, however, on limited conditions where the data is encrypted, they cannot be applied. In this work, we present a framework for measuring the level of engagement of group conversations and users, without reading the messages. Our framework creates an ensemble of interaction networks that represent the temporal evolution of the conversation, then, we apply the proposed Engagement Index (EI) for each interval of conversations to asses users' participation. Our results in five datasets from real-world WhatsApp Groups indicate that, based on the EI, it is possible to identify the most engaged users within a time interval, create rankings and group users according to their engagement and monitor their performance over time.

Outlier detection is a fundamental task for knowledge discovery in data mining, especially in the Big Data era. It aims to detect data items that deviate from the general pattern of a given data set. In this paper, we present a new outlier detection technique using tourist walks starting from each data sample and varying the memory size. Specifically, a data sample gets a higher outlier score if it participates in few tourist walk attractors, while it gets a low score if it participates in a large number of attractors. Experimental results on artificial and real data sets show good performance of the proposed method. In comparison to classical outlier detection methods, the proposed one shows the following salient features: (1) It finds out outliers by identifying the structure of the input data set instead of considering only physical features, such as distance, similarity or density. (2) It can detect not only external outliers as classical methods do, but also internal outliers staying among various normal data groups. (3) By varying the memory size, the tourist walks can characterize both local and global structures of the data set. (4) A parallel implementation is quite convenient due to the nature of large amount of independent walking of the algorithm. (5) The proposed method is a deterministic technique. Therefore, only one run is sufficient, in contrast to stochastic techniques, which require many runs. Moreover, in this work, we find, for the first time, that tourist walks can generate complex attractors in various crossing shapes. Such complex attractors reveal data structures in more details. Consequently, it can improve the outlier detection performance.

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.

Data and knowledge representation are fundamental concepts in machine learning. The quality of the representation impacts the performance of a learning model directly. Feature learning transforms or enhances raw data to structures that are effectively exploited by those methods. In recent years, several works have been using complex networks for data representation and analysis. However, no feature learning method has been proposed to enhance such category of representation. Here, we present an unsupervised feature learning mechanism that works on datasets with binary features. First, the dataset is mapped into a feature-sample network. Then, a multi-objective optimization process selects a set of new vertices to produce an enhanced version of the network. The new features depend on a nonlinear function of a combination of preexisting features. Effectively, the process projects the input data into a higher-dimensional space. To solve the optimization problem, we design two metaheuristics based on the lexicographic genetic algorithm and the improved strength Pareto evolutionary algorithm (SPEA2). We show that the enhanced network contains more useful information and can be exploited to improve the performance of machine learning methods. The advantages and disadvantages of each optimization strategy are discussed.

Extracting knowledge from time series analysis has been growing in importance and complexity over the last decade as the amount of stored data has increased exponentially. Considering this scenario, new data mining techniques have continuously developed to deal with such a situation. In this paper, we propose to study time series based on its topological characteristics, observed on complex networks generated from the time series data. Specifically, the aim of the proposed model is to create a trend detection algorithm for stochastic time series based on community detection and network walk observations. It is expected that the proposed model presents some advantages over traditional time series analysis, such as dimensionality reduction, use of hidden correlation on data and reinforcement learning as more data is added to the data set. Experimental results on the Bovespa index (Brazilian stock market) trend prediction shows that the proposed technique is promising.

In machine learning, traditional data classification techniques analyze only physical features of the input data (e.g., distance or distribution) in order to identify the main differences among them. This type of approach is referred to as low level classification. However, the human (animal) brain is able to perform not only low orders of learning, but it is also able to identify patterns according to the semantic meaning of the input data. Data classification that considers both physical attributes and also the pattern formation, is referred to as high level classification. Previous high level classification techniques require a low level technique to work together. Such an approach does not fully highlight the ability of feature extraction embedded in high-level schemes. In this paper, we propose a pure network- based high level classification technique which aims to identify the classes of new instances by detecting and comparing the impact that each one of them has on the topological structure of the network components, which represent each class of the input data set. Eight artificially generated data sets, along with other nine different real classification data sets, were used in order to test this technique, as well as to compare its performance with those obtained by nine traditional and well-known classification models. The results of these tests are very stimulating, indicating that the novel technique proposed in this work may have great potential for further development and application. Moreover, the peculiarity of the concept of pattern based classification provides a new general approach for raw data feature extraction.

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.

Accepted version of publication "NK Hybrid Genetic Algorithm for Clustering", published in IEEE Transactions on Evolutionary Computation

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.

The emergence of collective dynamics in neural networks is a mechanism of the animal and human brain for information processing. In this paper, we develop a computational technique using distributed processing elements in a complex network, which are called particles, to solve semisupervised learning problems. Three actions govern the particles' dynamics: generation, walking, and absorption. Labeled vertices generate new particles that compete against rival particles for edge domination. Active particles randomly walk in the network until they are absorbed by either a rival vertex or an edge currently dominated by rival particles. The result from the model evolution consists of sets of edges arranged by the label dominance. Each set tends to form a connected subnetwork to represent a data class. Although the intrinsic dynamics of the model is a stochastic one, we prove that there exists a deterministic version with largely reduced computational complexity; specifically, with linear growth. Furthermore, the edge domination process corresponds to an unfolding map in such way that edges "stretch" and "shrink" according to the vertex-edge dynamics. Consequently, the unfolding effect summarizes the relevant relationships between vertices and the uncovered data classes. The proposed model captures important details of connectivity patterns over the vertex-edge dynamics evolution, in contrast to the previous approaches, which focused on only vertex or only edge dynamics. Computer simulations reveal that the new model can identify nonlinear features in both real and artificial data, including boundaries between distinct classes and overlapping structures of data.

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.

High Level Classification seeks to identify class patterns based not only on physical features, such as distances among objects. For such, one of the recent approaches used is based on complex networks and its' topological properties. This approach has as advantage the capability to discriminate between highly complex class structures. However, from previous studies, it is known that a combination of high and low level classifiers is better than the use of the classifiers separately. In this work, we propose the use of a stacking procedure to combine such classifiers. The main advantage, compared to the current procedure, is the removal of critical parameters, which have major influence on the results. Also, we propose two new measures to capture different global patterns from the classes complex networks. We performed experiments on five UCI datasets and a subset of the Million song Dataset. Our results indicate that the use of stacking is capable of obtaining equal or better results when compared to optimizing the critical parameters by cross-validation.

Graph-based algorithms have been successfully applied in machine learning and data mining tasks. A simple but, widely used, approach to build graphs from vector-based data is to consider each data instance as a vertex and connecting pairs of it using a similarity measure. Although this abstraction presents some advantages, such as arbitrary shape representation of the original data, it is still tied to some drawbacks, for example, it is dependent on the choice of a pre-defined distance metric and is biased by the local information among data instances. Aiming at exploring alternative ways to build graphs from data, this paper proposes an algorithm for constructing a new type of graph, called Attribute-based Decision Graph-AbDG. Given a vector-based data set, an AbDG is built by partitioning each data attribute range into disjoint intervals and representing each interval as a vertex. The edges are then established between vertices from different attributes according to a pre-defined pattern. Classification is performed through a matching process among the attribute values of the new instance and AbDG. Moreover, AbDG provides an inner mechanism to handle missing attribute values, which contributes for expanding its applicability. Results of classification tasks have shown that AbDG is a competitive approach when compared to well-known multiclass algorithms. The main contribution of the proposed framework is the combination of the advantages of attribute-based and graph-based techniques to perform robust pattern matching data classification, while permitting the analysis the input data considering only a subset of its attributes.

The use of good evaluation functions is essential when evolutionary algorithms are employed for clustering. The NK internal clustering validation measure is proposed for hard partitional clustering. The evaluation function is composed of N subfunctions, where N is the number of objects in the dataset. Each subfunction is influenced by a group of K+1 objects. By using neighbourhood relations among connected small groups, density-based regions can be identified. The NK internal clustering validation measure allows the application of partition crossover (PX). PX for hard partitional clustering is also proposed in this work. By using PX, the evaluation function can be decomposed in q partial evaluations. As a consequence, PX deterministically finds the best of 2q possible offspring at the cost of evaluating 2 solutions. In the experiments, the application of PX resulted in a high number of successful recombinations. It was able to improve partitions defined by the best parents.

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 class (cluster) formation process of machine learning techniques, data instances are usually assumed to have equal relevance. However, it is frequently not true. Such a situation is more typical in semi-supervised learning since we have to understand the data structure of both labeled and unlabeled data at the same time. In this paper, we investigate the organizational heterogeneity of data in semi-supervised learning using graph representation. This is because graph is a natural choice to characterize relationship between any pair of nodes or any pair of groups of nodes, consequently, strategical location of each node or each group of nodes can be determined by graph measures. Specifically, two issues are addressed: (1) We propose an adaptive graph construction method, we call AdaRadius, considering the heterogeneity of local interacting structure among nodes. As a result, it presents several interesting properties, namely adaptability to data density variations, low dependency on parameters setting, and reasonable computational cost, for both pool based and incremental data. (2) Moreover, we present heuristic criteria for selecting representative data samples to be labeled. Experimental study shows that selective labeling usually gets better classification results than random labeling. To our knowledge, it still lacks investigation on both issues up to now, therefore, our approach presents an important step toward the data heterogeneity characterization not only in semi-supervised learning, but also in general machine learning.

Real scenes are composed of multiple points possessing distinct characteristics. Selectively, only part of the scene undergoes scrutiny at a time, and the mechanism responsible for this task is named selective visual attention. Spatial location with the highest contrast might highlight from scene reaching level of awareness (bottom-up attention). On the other hand, attention may also be voluntarily directed to a particular object in the scene (object-based attention), which requires the recognition of a specific target (top-down modulation). In this paper, a new visual selection model is proposed, which combines both early visual features and object-based visual selection modulations. The possibility of the modulation regarding specific features enables the model to be applied to different domains. The proposed model integrates three main mechanisms. The first handles the segmentation of the scene allowing the identification of objects. In the second one, the average of saliency of each object is computed, which provides the modulation of the visual attention for one or more features. Finally, the third builds the object saliency map, which highlights the salient objects in the scene. We show that top-down modulation has a stronger effect than bottom-up saliency when a memorized object is selected, and this evidence is clearer in the absence of any bottom-up clue. Experiments with synthetic and real images are conducted, and the obtained results demonstrate the effectiveness of the proposed approach for visual selection. (C) 2015 Elsevier B.V. All rights reserved.

We present a biologically inspired model for transductive semi-supervised learning tasks. Specifically, this model consists of a set of particles that walk and compete in a complex network. From an input dataset, similarities between labeled and unlabeled data points derive a network representation. As particles walk the network, they compete to dominate the edges. Over the process, particles can become inactive, and, to compensate, labeled vertices will feed new particles to the system. Resulted from the model simulation, we analyze sets of edges arranged by their label dominance. Each set forms a subnetwork that is used to classify connected vertices. Our computer simulations on artificial and real datasets show that this technique can classify nonlinearly formed data and detect vertices of different classes in overlapping regions.

Many interactive image processing approaches are based on semi-supervised learning, which employ both labeled and unlabeled data in its training process. In the interactive image segmentation problem, a human specialist labels some pixels of an object while the semi-supervised algorithm labels the remaining pixels of the segment. The particle competition and cooperation model is a recent graph-based semi-supervised learning approach. It employs particles walking in a graph to classify the data items corresponding to graph nodes. Each particle group aims to dominate most unlabeled nodes, spreading their label, and preventing enemy particles invasion. In this paper, the particle competition and cooperation model is extended to perform interactive image segmentation. Each image pixel is converted into a graph node, which is connected to its nearest neighbors according to their visual features and location in the original image. Labeled pixel generates particles that propagate their label to the unlabeled pixels. The particle model also takes the contributions from the adjacent pixels to classify less confident labeled pixels. Computer simulations are performed on real-world images, including images from the Microsoft GrabCut dataset, which allows a straightly comparison with other techniques. The segmentation results show the effectiveness of the proposed approach.

Missing attribute values is a recurrent problem in data mining and machine learning. Although there are plenty of techniques to handle this problem, most of them are too simplistic to provide a good estimation for absent attribute values. A very active research area focuses on solving the missing attribute value problem via imputation methods, which replaces missing data with substituted values. This paper proposes a new imputation method which uses a special graph named Complete p-Partite Attribute-based Decision Graphs (CpP-AbDG) to estimate, in a consistent and plausible way, the missing values. The graph is built by considering the range of each attribute that describes the data divided into sub-intervals; sub-intervals are approached as the vertices of a graph. Edges are then established between pairs of different vertices, provided they do not related to the same attribute. The edges and vertices are finally assigned a weight, based on distributions of the classes. The resulting CpP-AbDG has shown to be a suitable and informative data structure for finding the proper interval in which a missing attribute value should lie, taking into account all the attributes that describe the data. Results comparing the proposed approach to classical ones in an computational environment that considers classification problems as an evaluation criteria, show the potential of the method.

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.

We present a nature-inspired semi-supervised learning technique based on the flocking formation of certain living species like birds and fishes. Each data item is treated as an individual in the flock. Starting from random directions, each data item moves according to its surrounding items, by getting closer to them (but not too much close) and taking the same direction of motion. Labeled items play special roles, ensuring that data from different classes will belong to different, distant flocks. Experiments on both artificial and benchmark datasets were performed and show its classification accuracy. Despite the rich behavior, we argue that this technique has a sub-quadratic asymptotic time complexity, thus being feasible to be used on large datasets. In order to achieve such performance, a space-partitioning technique is introduced. We also argue that the richness behind this dynamic, self-organizing model is quite robust and may be used to do much more than simply propagating the labels from labeled to unlabeled data. It could be used to determine class overlapping, wrong labeling, etc.

Non-stationary data can be characterized as data having a distribution that changes over time. It is well-known that most successful machine learning algorithms are based on stationary data i.e., data that are assumed to have a fixed distribution (although unknown, in most cases). Non-stationary classification problems require the induced classifiers to be flexible enough to learn or adapt themselves to reflect the changes on data distribution over time; this can be a hard task, taking into account that changes that may happen are not usually known in advance. Although there are several proposals in the literature that deal with non-stationary data, none of them deal with missing attribute values, a common problem in real applications. This paper proposes an ensemble of classifiers for non-stationary environments that (1) uses a new graph structure for representing data known as Complete P-partite Attribute-based Decision Graph - CPp-AbDG; (2) handles data described by heterogeneous attributes (numeric and categorical) and (3) handles missing attribute values. Experiments in non-stationary environments show evidence of the strength of the CPp-AbDG representation as well as the potentiality of the proposed ensemble approach.

Graph-based representation has been successfully used to support various machine learning and data mining algorithms. The learning algorithms strongly rely on the algorithm employed for constructing the graph from input data, given as a set of vector-based patterns. A popular way to build such graphs is to treat each data pattern as a vertex; vertices are then connected according to some similarity measure, resulting in an structure known as data graph. In this paper we propose a new type of data graph, focused on data attributes, named Attribute-based Decision Graph - AbDG, suitable for supervised multiclass classification tasks. The input data for constructing an AbDG is a set of data-vectors (patterns), that can be described by either type of attributes (numeric, categorical or both). Also, algorithms for constructing such graphs and using them in classification tasks are described. An AbDG can be associated to a classifying procedure approached as a graph matching process, where the sub-graph representing a new pattern is matched against the AbDG. The proposed approach has been experimentally evaluated on classification tasks in twenty knowledge domains and the results are competitive when compared to those of two well-known classification methods (C4.5 and Multi-Interval ID3).

Network-based Semi-Supervised Learning (NBSSL) propagates labels in networks constructed from the original vector-based data sets taking advantage of the network topology. However, the NBSSL classification performance often varies according to the representativeness of the labeled data instances. Herein, we address this issue. We adopt heuristic criteria for selecting data items for manual labeling based on complex networks centrality measures. The numerical analysis are performed on Girvan and Newman homogeneous networks and Lancichinetti-Fortunato-Radicchi heterogeneous networks. Counterintuitively, we found that the highly connective nodes (hubs) are usually not representative, in the sense that random samples performs as well as them or even better. Other than expected, nodes with high clustering coefficient are good representatives of the data in homogeneous networks. On the other hand, in heterogeneous networks, nodes with high betweenness are the good representatives. A high clustering coefficient means that the node lies in a much connected motif (clique) and a high betweenness means that the node lies interconnecting modular structures. Moreover, aggregating the complex networks measures through Principal Components Analysis, we observed that the second principal component (Z2) exhibits potentially promising properties. It appears that Z2is able to extract discriminative characteristics allowing finding good representatives of the data. Our results reveal that the performance of the NBSSL can be significantly improved by finding and labeling representative data instances.

In this paper a new neural network system for composition of melodies is proposed. The Long Short-Term Memory (LSTM) neural network is adopted as the neural network model. We include an independent melody as an additional input in order to provide an inspiration source to the network. This melody is given by a chaotic composition algorithm and works as an inspiration to the network enhancing the subjective measure of the composed melodies. As the chaotic system we use the Hénon map with two variables, which are mapped to pitch and rhythm. We adopt a measure to conduct the degree of melodiousness (Euler's gradus suavitatis) of the output melody, which is compared with a reference value. Varying a specific parameter of the chaotic system, we can control the complexity of the chaotic melody. The system runs until the degree of melodiousness falls within a predetermined range.

Traditional data classification is based only on physical features of input data. They are called low level classification. Data classification by considering not only physical attributes but also pattern formation is denominated high level classification. In this paper, we propose a new technique that performs high level classification by extracting information of networks constructed from the input data. Specifically, we calculate the network entropies before and after the insertion of a data item to be classified. Then, we classify it as belonging to the class which results in the largest increase of the entropy measures. We show that the proposed method can execute classification tasks according to both similarity and pattern formation of input data to reach good results in the experiments with artificial and real data sets. In summary, our technique can calculate how significant a data item is for each class performing a new way to classify data.

The cerebral cortex is a complex network. It contains billions of neurons divided in spatial and functional clusters to perform different tasks. It also operates with complex dynamics such as periodic and chaotic ones. It has been shown that chaotic neural networks are more efficient than conventional recurrent neural networks in avoiding spurious memory. Inspired by the fact that the cerebral cortex has specific groups of cells, in this paper we investigate the dynamic of a recurrent neural network where neurons are coupled in such a way that form communities of a complex network. Also, we generate an asymmetric weight matrix placing pattern cycles during learning. Such a learning rule provides a natural periodic behavior in a fully connected network. Community structure breaks the connections up, forcing chaos to emerge. Our study shows that chaotic behavior rises for a high fragmentation degree in either just one community with sparse connections or several communities with few inter-community connections. For the latter case, we also show that the neural network can hold chaotic dynamic and a high value of modularity measure at the same time. These findings provide an alternative way to design dynamical neural networks to perform pattern recognition tasks exploiting periodic and chaotic dynamics.

Recently, it is observed that, in a single neural model, phase noise (time-varying signal phase) arising from an external stimulating signal can induce regular spiking activities even if the signal is subthreshold. In addition, it is also uncovered that there exists an optimal phase noise intensity at which the spiking rhythm coincides with the frequency of the subthreshold signal, resulting in a phase-noise-induced resonance phenomenon. However, neurons usually do not work alone, but are connected in the form of arrays or blocks. Therefore, we study the spiking activity induced by phase noise in arrays of globally and locally coupled excitable neural models. We find that there also exists an optimal phase noise intensity for generating large neural response and such an optimal value is significantly decreased compared to an isolated single neuron case, which means the detectability in response to the subthreshold signal of neurons is sharply improved because of the coupling. In addition, we reveal two new resonance behaviors in the neuron ensemble with the presence of phase noise: there exist optimal values of both coupling strength and system size, where the coupled neurons generate regular spikes under subthreshold stimulations, which are called as coupling strength and system size resonance, respectively. Finally, the dependence of phase-noise-induced resonance on signal frequency is also examined.

There are several real situations in which it is useful to have a system able to detect a specific target or a salient object and its localization in a given image in autonomous way. To guide the attention based on known characteristics of an object and primitive information of the image is not a trivial task for visual attention. Several works about visual attention have been developed, which focused in bottom-up or top-down in an isolate manner. We propose in this work a model of visual attention combining characteristics bottom-up and top-down. The proposed model is composed of four components: the training and recognition of known objects, the object segmentation of the input image, the self-organizing of information top-down and bottom-up in a single map and a network of neurons with excitatory connections and inhibitory connections to generate the map of salient attribute for the location of salient objects. Thanks to this combination it was possible to detect, to identify and to locate the salient objects of the image. Several tests have been applied to synthetic images to verify the viability of the model as a mechanism of selection of objects as a part of a visual attention system. The results demonstrate the effectiveness of the model.

Concept drift is a problem of increasing importance in machine learning and data mining. Data sets under analysis are no longer only static databases, but also data streams in which concepts and data distributions may not be stable over time. However, most learning algorithms produced so far are based on the assumption that data comes from a fixed distribution, so they are not suitable to handle concept drifts. Moreover, some concept drifts applications requires fast response, which means an algorithm must always be (re)trained with the latest available data. But the process of labeling data is usually expensive and/or time consuming when compared to unlabeled data acquisition, thus only a small fraction of the incoming data may be effectively labeled. Semi-supervised learning methods may help in this scenario, as they use both labeled and unlabeled data in the training process. However, most of them are also based on the assumption that the data is static. Therefore, semi-supervised learning with concept drifts is still an open challenge in machine learning. Recently, a particle competition and cooperation approach was used to realize graph-based semi-supervised learning from static data. In this paper, we extend that approach to handle data streams and concept drift. The result is a passive algorithm using a single classifier, which naturally adapts to concept changes, without any explicit drift detection mechanism. Its built-in mechanisms provide a natural way of learning from new data, gradually “forgetting” older knowledge as older labeled data items became less influent on the classification of newer data items. Some computer simulation are presented, showing the effectiveness of the proposed method.

Wireless Sensor Networks (WSN) are a special kind of ad-hoc networks that is usually deployed in a monitoring field in order to detect some physical phenomenon. Due to the low dependability of individual nodes, small radio coverage and large areas to be monitored, the organization of nodes in small clusters is generally used. Moreover, a large number of WSN nodes is usually deployed in the monitoring area to increase WSN dependability. Therefore, the best cluster head positioning is a desirable characteristic in a WSN. In this paper, we propose a hybrid clustering algorithm based on community detection in complex networks and traditional K-means clustering technique: the QK-Means algorithm. Simulation results show that QK-Means detect communities and sub-communities thus lost message rate is decreased and WSN coverage is increased.

In this paper, we present a method for determining overlapping cluster structures in the network using a particle competition model. Specifically, several particles walk in the network and compete with each other to occupy as many nodes as possible, while attempting to reject intruder particles. The overlapping nodes in the input data are uncovered by using the domination level information generated by the competition process itself. In this way, the detection procedure is already embedded in the model, which in turn has low computational complexity. Computer simulations reveal that this overlapping index works well in real-world data sets. Finally, an application on handwritten data clustering is provided and high clustering accuracies are obtained.

Semi-supervised learning is one of the important topics in machine learning, concerning with pattern classification where only a small subset of data is labeled. In this paper, a new network-based (or graph-based) semi-supervised classification model is proposed. It employs a combined random-greedy walk of particles, with competition and cooperation mechanisms, to propagate class labels to the whole network. Due to the competition mechanism, the proposed model has a local label spreading fashion, i.e., each particle only visits a portion of nodes potentially belonging to it, while it is not allowed to visit those nodes definitely occupied by particles of other classes. In this way, a “divide-and-conquer” effect is naturally embedded in the model. As a result, the proposed model can achieve a good classification rate while exhibiting low computational complexity order in comparison to other network-based semi-supervised algorithms. Computer simulations carried out for synthetic and real-world data sets provide a numeric quantification of the performance of the method.

The ability to transmit and amplify weak signals is fundamental to signal processing of artificial devices in engineering. Using a multilayer feedforward network of coupled double-well oscillators as well as Fitzhugh-Nagumo oscillators, we here investigate the conditions under which a weak signal received by the first layer can be transmitted through the network with or without amplitude attenuation. We find that the coupling strength and the nodes' states of the first layer act as two-state switches, which determine whether the transmission is significantly enhanced or exponentially decreased. We hope this finding is useful for designing artificial signal amplifiers.

Competitive learning is an important machine learning approach which is widely employed in artificial neural networks. In this paper, we present a rigorous definition of a new type of competitive learning scheme realized on large-scale networks. The model consists of several particles walking within the network and competing with each other to occupy as many nodes as possible, while attempting to reject intruder particles. The particle's walking rule is composed of a stochastic combination of random and preferential movements. The model has been applied to solve community detection and data clustering problems. Computer simulations reveal that the proposed technique presents high precision of community and cluster detections, as well as low computational complexity. Moreover, we have developed an efficient method for estimating the most likely number of clusters by using an evaluator index that monitors the information generated by the competition process itself. We hope this paper will provide an alternative way to the study of competitive learning..

Semisupervised learning is a machine learning approach that is able to employ both labeled and unlabeled samples in the training process. In this paper, we propose a semisupervised data classification model based on a combined random-preferential walk of particles in a network (graph) constructed from the input dataset. The particles of the same class cooperate among themselves, while the particles of different classes compete with each other to propagate class labels to the whole network. A rigorous model definition is provided via a nonlinear stochastic dynamical system and a mathematical analysis of its behavior is carried out. A numerical validation presented in this paper confirms the theoretical predictions. An interesting feature brought by the competitive-cooperative mechanism is that the proposed model can achieve good classification rates while exhibiting low computational complexity order in comparison to other network-based semisupervised algorithms. Computer simulations conducted on synthetic and real-world datasets reveal the effectiveness of the model.

Traditional supervised data classification considers only physical features (e.g., distance or similarity) of the input data. Here, this type of learning is called low level classification. On the other hand, the human (animal) brain performs both low and high orders of learning and it has facility in identifying patterns according to the semantic meaning of the input data. Data classification that considers not only physical attributes but also the pattern formation is, here, referred to as high level classification. In this paper, we propose a hybrid classification technique that combines both types of learning. The low level term can be implemented by any classification technique, while the high level term is realized by the extraction of features of the underlying network constructed from the input data. Thus, the former classifies the test instances by their physical features or class topologies, while the latter measures the compliance of the test instances to the pattern formation of the data. Our study shows that the proposed technique not only can realize classification according to the pattern formation, but also is able to improve the performance of traditional classification techniques. Furthermore, as the class configuration's complexity increases, such as the mixture among different classes, a larger portion of the high level term is required to get correct classification. This feature confirms that the high level classification has a special importance in complex situations of classification. Finally, we show how the proposed technique can be employed in a real-world application, where it is capable of identifying variations and distortions of handwritten digit images. As a result, it supplies an improvement in the overall pattern recognition rate.

In this work, an Elman recurrent neural network is used for automatic musical structure composition based on the style of a music previously learned during the training phase. Furthermore, a small fragment of a chaotic melody is added to the input layer of the neural network as an inspiration source to attain a greater variability of melodies. The neural network is trained by using the BPTT (back propagation through time) algorithm. Some melody measures are also presented for characterizing the melodies provided by the neural network and for analyzing the effect obtained by the insertion of chaotic inspiration in relation to the original melody characteristics. Specifically, a similarity melodic measure is considered for contrasting the variability obtained between the learned melody and each one of the composite melodies by using different quantities of inspiration musical notes.

Networks are powerful representations for many complex systems, where nodes represent elements of the system and edges represent connections between them. Consensus problems in coupled agents have already been studied in complex networks. This paper explores the use of the consensus time in the presence of a leader as distance measure on complex networks. In this case, the distance between two nodes in the network is characterized by the time that one of them takes to reach a stationary state with the other node being pinned. A new technique for community detection of complex networks has been developed based on the proposed distance measure. The method has been tested with various networks and promising results have been obtained.

Complex network provides a general scheme for machine learning. In this paper, we propose a competitive learning mechanism realized on large scale networks, where several particles walk in the network and compete with each other to occupy as many nodes as possible. Each particle can perform a random walk by choosing any neighbor to visit, a deterministic walk by choosing to visit the node with the highest domination, or a combination of them. A computational complexity analysis is developed of the proposed algorithm. Computer simulations performed on several real-world data sets, including a large scale data set, reveal attractive results when the model is applied for data clustering problems.

Identifying outlier nodes is an important task in complex network mining. In this paper, we analyze the problem of identifying outliers in a network structure and propose an outlier measure by using the random walk distance measure and the dissimilarity index between pairs of vertices. Our method determines a “view” to the whole network for each node and infers that outliers are those nodes whose views differ significantly from majority of the nodes. Usually, outlier is detected by applying a specific criteria, for example, the farthest ones from the central node. Consequently, only one type of outliers satisfying the predefined criteria can be determined. On the other hand, our method incorporates both local and global information of the network due to random walk feature and can give more general outlier detection results. We have applied the method to artificial and real networks and some interesting results have been obtained.

In machine learning study, semi-supervised learning has received increasing interests in the last years. It is applied to classification problems where only a small portion of the data points is labeled. In these situations, the reliability of these labels is extremely important because it is common to have mislabeled samples in a data set and these may propagate their wrong labels to a large portion of the data set, resulting in major classification errors. In spite of its importance, wrong label propagation in semi-supervised learning has received little attention from researchers. In this paper we propose a particle walk semi-supervised learning method with both competitive and cooperative mechanisms. Then we study error propagation by applying the proposed model in modular networks. We show that the model is robust against mislabeled samples and it can produce good classification results even in the presence of considerable proportion of mislabeled data. Moreover, our numerical analysis uncover a critical point of mislabeled subset size, below which the network is free of wrong label contamination, but above which the mislabeled samples start to propagate their labels to the rest of the network. These studies have practical importance to design secure and robust machine learning techniques.

Semi-Supervised Learning (SSL) is a machine learning research area aiming the development of techniques which are able to take advantage from both labeled and unlabeled samples. Additionally, most of the times where SSL techniques can be deployed, only a small portion of samples in the data set is labeled. To deal with such situations in a straightforward fashion, in this paper we introduce a semi-supervised learning approach based on neuronal synchrony in a network of coupled integrate-and-fire neurons. For that, we represent the input data set as a graph and model each of its nodes by an integrate-and-fire neuron. Thereafter, we propagate the class labels from the seed samples to unlabeled samples through the graph by means of the emerging synchronization dynamics. Experimentations on synthetic and real data show that the introduced technique achieves good classification results regardless the feature space distribution or geometrical shape.

Chaotic phase synchronization among coupled oscillators is a phenomenon of interest in many physical and engineering systems. It has also been observed in biological systems, where groups of different functional units interact with each other in order to produce coherent behaviors in higher levels. While biological systems have facility to capture salient object(s) in a given scene, visual selection is still a challenging task to artificial visual systems. In this paper, a visual selection mechanism based on chaotic phase synchronization is proposed. Oscillators representing the salient object in a given scene are phase synchronized, while no synchronization is observed for background objects. In this way, the salient object is highlighted. Due to the modeling by phase synchronization instead of complete synchronization, the proposed model is robust, biologically inspired and good simulation results were achieved.

Attention is a critical mechanism for visual scene analysis. By means of attention, it is possible to break down the analysis of a complex scene to the analysis of its parts through a selection process. Empirical studies demonstrate that attentional selection is conducted on visual objects as a whole. We present a neurocomputational model of object-based selection in the framework of oscillatory correlation. By segmenting an input scene and integrating the segments with their conspicuity obtained from a saliency map, the model selects salient objects rather than salient locations. The proposed system is composed of three modules: a saliency map providing saliency values of image locations, image segmentation for breaking the input scene into a set of objects, and object selection which allows one of the objects of the scene to be selected at a time. This object selection system has been applied to real images and the simulation results show its effectiveness.

Data clustering is an important technique to extract and understand relevant information in large data sets. In this paper, a clustering algorithm based on graph theoretic models and community detection in complex networks is proposed. Two steps are involved in this processing: The first step is to represent input data as a network and the second one is to partition the network into subnetworks producing data clusters. In the network partition stage, each node has a randomly assigned initial angle and it is gradually updated according to its neighbors angle agreement. Finally, a stable state is reached and nodes belonging to the same cluster have similar angles. This process is repeated, each time a cluster is chosen and results in an hierarchical divisive clustering. Simulation results show two main advantages of the algorithm: the ability to detect clusters in different shapes, densities and sizes and the ability to generate clusters with different refinement degrees. Besides of these, the proposed algorithm presents high robustness and efficiency in clustering.

In this paper a visual selection mechanism based on the pulse-coupled neural network is proposed for discriminating one object among others in a given visual scene by delivering the focus of attention to the salient object. In comparison to other visual selection approaches, this model presents at least two new features. First, it is able to highlight not only objects or components of objects in simple forms, but also objects in complex forms, including those that are non-linear separable, and second it provides a shifting mechanism from one previously selected object to another. Computer simulations are performed and the results show that the proposed model is promising as a selection mechanism embedded in a visual attention system.

Self-organization is a highly observed process of brain to solve several cognitive tasks and many artificial systems based on this mechanism have been developed. In this paper, a self-organized dynamical model is proposed for pixel clustering. This approach employs a network consisting of interacting elements with each representing a pixel value and receiving attractions from other elements within a certain neighborhood. Those attractions, determined by a predefined similarity measure, drive the elements to converge to their corresponding cluster center. With this model, neither the number of pixel clusters nor the initial guessing of cluster centers is required. Computer simulations for clustering of real images are performed to show the efficiency of the proposed approach.

This paper presents a performance comparison of various methods for associative memory design using cellular neural networks (CNNs). Even though there have been an increasing interest in such kind of application for CNNs, there is no proper comparison of performance among the available methods in the literature. This paper reviews methods of associative memory design based on CNNs, and provides comparative performance analyses of these approaches.

In this paper, a Visual Selection Mechanism based on a lattice of coupled chaotic Wilson-Cowan oscillators is proposed. The oscillators representing each object in a given visual scene are synchronized to produce a chaotic trajectory. Cooperation and competition mechanisms are also introduced to accelerate oscillating frequency of the salient object as well as to slow down other objects in the same scene. The model can not only discriminate each object among others in a given visual scene, but also deliver the focus of attention to the salient object. In comparison to other visual selection approaches, this model presents at least two new features. First, it is able to highlight objects in complex forms, including those that are non-linear separable. Second, oscillators representing the salient object will jump from chaotic phase to periodic phase. This behavior matches well to biological experiments on pattern recognition of rabbit. Computer simulations are performed and the results show that the proposed model is promising as a Selection Mechanism embedded in a Visual Attention System.

Traditional data clustering techniques present difficulty in determination of clusters of arbitrary forms. On the other hand, graph theoretic methods seek topological orders among input data and, consequently, can solve the above mentioned problem. In this paper, we present an improved graph theoretic model for data clustering. The clustering process of this model is composed of two steps: network formation by using input data and hierarchical network partition to obtain clusters in different scales. Our network formation method always produces a connected graph with densely linked nodes within a community and sparsely linked nodes among different communities. The community detection technique used here has the advantage that it is completely free from physical distances among input data. Consequently, it is able to discover clusters of various forms correctly. Computer simulations show the promising performance of the model.

In this paper, a network of coupled chaotic maps for pixel clustering is proposed. Time evolutions of chaotic maps in the network corresponding to a pixel cluster are synchronized with each other. Those synchronized trajectories are desynchronized with respect to the time evolutions of chaotic maps corresponding to other pixel clusters in the same image. A pixel motion mechanism is also introduced, which makes each group of pixels more compact and, consequently, makes the model robust enough to classify ambiguous pixels. Another feature of the proposed model is that the number of pixel clusters does not need to be previously known.

Human perception is a complex nonlinear dynamics. On the one hand it is periodic dynamics and on the other hand it is chaotic. Thus, we wish to propose a hybrid-the spatial chaotic dynamics for the associative recall to retrieve patterns, similar to Walter Freeman's discovery, and the fixed point dynamics for memory stage, similar to Hopfield and Grossberg's discoveries. In this model, each neuron in the network could be a chaotic map, whose phase space is divided into two states: one is periodic dynamic state with period-V, which is used to represent a V-value retrieved pattern; another is chaotic dynamic state. Firstly, patters are stored in the memory by fixed point learning algorithm. In the retrieving process, all neurons are initially set in the chaotic region. Due to the ergodicity property of chaos, each neuron will approximate the periodic points covered by the chaotic attractor at same instants. When this occurs, the control is activated to drive the dynamic of each neuron to their corresponding stable periodic point. Computer simulations confirm the theoretical prediction.

In this paper, a network of coupled chaotic maps for multi-scale image segmentation is proposed. Time evolutions of chaotic maps that correspond to a pixel cluster are synchronized with one another, while this synchronized evolution is desynchronized with respect to time evolution of chaotic maps corresponding to other pixel clusters in the same image. The number of pixel clusters is previously unknown and the adaptive pixel moving technique introduced in the model makes it robust enough to classify ambiguous pixels.

In this paper, a network of coupled chaotic maps for multi-scale image segmentation is proposed. Time evolutions of chaotic maps that correspond to a pixel cluster are synchronized with one another, while this synchronized evolution is desynchronized with respect to time evolution of chaotic maps corresponding to other pixel clusters in the same image. The number of pixel clusters is previously unknown and the adaptive pixel moving technique introduced in the model makes it robust enough to classify ambiguous pixels.

In this paper, a computational model for scene segmentation based on a network of dynamically coupled chaotic maps is proposed. Time evolutions of chaotic maps that correspond to an object in the given scene are synchronized with one another, while this synchronized evolution is desynchronized with respect to time evolution of chaotic maps corresponding to other objects in the scene. In this model, the coupling range of each active element increases dynamically according to predefined rules until a saturated state is achieved, i.e., locally coupled chaotic maps corresponding to an object in the initial state will be coupled globally in the final state. Consequently, the advantage of both global coupling and local coupling are incorporated in a single scheme. In comparison to continuous models, this proposed model is suitable for computational implementation. Another significant benefit is that the good performance and transparent dynamics of the model are obtained by utilizing one-dimensional chaotic map instead of complex neuron as each element.