arXiv.itsnick.co

an independent field guide to 23 million papers from arXiv & OpenAlex — not affiliated with arXiv.org

Author

Addisson Salazar

41 papers · 2005–2023

2005: 22006: 12007: 22008: 42009: 12010: 22011: 22012: 32013: 12014: 32015: 22016: 22017: 42018: 32019: 32020: 22021: 22022: 02023: 2
2005peak 4 in 20082023

Frequent co-authors

Papers

showing 1–25 of 41 RSS

Generative Adversarial Networks and Markov Random Fields for oversampling very small training sets

Addisson Salazar, Luis Vergara, Gonzalo Safont

Artificial Intelligence · Expert Systems with Applications · · cited 112× · DOI ↗ · more like this ·

In this work, we propose a new method for oversampling the training set of a classifier, in a scenario of extreme scarcity of training data. It is based on two concepts: Generative Adversarial Networks (GAN) and vector Markov Random Field (vMRF). Thus, the generative block of GAN uses the vMRF model to synthesize surrogates by the Graph Fourier Transform. Then, the discriminative block implements a linear discriminant on features measuring clique similarities between the synthesized and the original instances. Both blocks iterate until the linear discriminant cannot discriminate the synthetic from the original instances. We have assessed the new method, called Generative Adversarial Network Synthesis for Oversampling (GANSO), with both simulated and real data in experiments where the classifier is to be trained with just 3 or 5 instances. The applications consisted of classification of…

Fusion of Scores in a Detection Context Based on Alpha Integration

Antonio Soriano, Luis Vergara, Ahmed Bouziane +1 more

Artificial Intelligence · Neural Computation · · cited 58× · DOI ↗ · more like this ·

We present a new method for fusing scores corresponding to different detectors (two-hypotheses case). It is based on alpha integration, which we have adapted to the detection context. Three optimization methods are presented: least mean square error, maximization of the area under the ROC curve, and minimization of the probability of error. Gradient algorithms are proposed for the three methods. Different experiments with simulated and real data are included. Simulated data consider the two-detector case to illustrate the factors influencing alpha integration and demonstrate the improvements obtained by score fusion with respect to individual detector performance. Two real data cases have been considered. In the first, multimodal biometric data have been processed. This case is representative of scenarios in which the probability of detection is to be maximized for a given probability…

Semi-Supervised Learning For Imbalanced Classification Of Credit Card Transaction

Addisson Salazar, Gonzalo Safont, Luis Vergara

Artificial Intelligence · · cited 49× · DOI ↗ · more like this ·

Success in supervised learning is constrained by availability of an adequate labeled data sample for training. The problem of a complete labeling of every data of the training dataset can be alleviated allowing semi-complete labeling in a way so called semi-supervised learning. In this paper, we investigate the performance of semi-supervised learning in imbalanced classification problems. Augmentation of the class of limited data is applied for lowering the variance of the estimate using a data subrogation method. We analyze the effect of this data augmentation in several simulated and experimental scenarios of a challenging application: automatic credit card fraud detection. The relationships among different semi-supervision and sample augmentation ratios in this application are discussed in terms of receiver operating characteristic curves and business key performance indicators.

A case study of knowledge discovery on academic achievement, student desertion and student retention

Addisson Salazar, J. Gosalbez, I. Bosch +2 more

Computer Science Applications · · cited 46× · DOI ↗ · more like this ·

This paper presents an applied research study on knowledge discovery based on academic data analysis. The main objectives of this research study were to get knowledge about academic achievement success and failure, student retention and student desertion. Automatic clustering and decision rule data mining techniques were used. The application of the C-mean algorithm on statistically homogeneous data subsets provided a group of clusters, which have been qualitatively described. Using selected clusters a decision rule study based on C4.5 algorithm generated a set of decision rules for the four research topics of the study. These decision rules have been interpreted, finding global content and specific content knowledge. The findings of the research study have been well evaluated by experts in aspects such as validity, novelty, simplicity and utility. In the last stage of the knowledge…

Surrogate techniques for testing fraud detection algorithms in credit card operations

Addisson Salazar, Gonzalo Safont, Luis Vergara

Artificial Intelligence · · cited 38× · DOI ↗ · more like this ·

Banks collect large amount of historical records corresponding to millions of credit cards operations, but, unfortunately, only a small portion, if any, is open access. This is because, e.g., the records include confidential customer data and banks are afraid of public quantitative evidence of existing fraud operations. This paper tackles this problem with the application of surrogate techniques to generate new synthetic credit card data. The quality of the surrogate multivariate data is guaranteed by constraining them to have the same covariance, marginal distributions, and joint distributions as the original multivariate data. The performance of fraud detection algorithms (in terms of receiver operating characteristic (ROC) curves) using a varying proportion of real and surrogate data is tested. We demonstrate the feasibility of surrogates in a real scenario considering very low false…

Automatic credit card fraud detection based on non-linear signal processing

Addisson Salazar, Gonzalo Safont, Antonio Soriano +1 more

Artificial Intelligence · · cited 34× · DOI ↗ · more like this ·

Fraud detection is a critical problem affecting large financial companies that has increased due to the growth in credit card transactions. This paper presents a new method for automatic detection of frauds in credit card transactions based on non-linear signal processing. The proposed method consists of the following stages: feature extraction, training and classification, decision fusion, and result presentation. Discriminant-based classifiers and an advanced non-Gaussian mixture classification method are employed to distinguish between legitimate and fraudulent transactions. The posterior probabilities produced by classifiers are fused by means of order statistical digital filters. Results from data mining of a large database of real transactions are presented. The feasibility of the proposed method is demonstrated for several datasets using parameters derived from receiver…

Design and implementation of transmitter and receiver filters with periodic coefficient nulls for digital systems

Addisson Salazar, Victor B. Lawrence

Signal Processing · · cited 34× · DOI ↗ · more like this ·

The problem of designing nonrecursive digital filters to perform baseband and passband pulse shaping in digital systems amounts to determining a discrete sequence with preassigned zero crossings. This approach yields a filter frequency response which permits minimal stop band noise enhancement. Several methods exist for designing transmitter and receiver filters for digital systems. An interesting technique using a minimax criterion in the design procedure employs linear programming techniques in the frequency domain, in addition to ensuring periodic nulls in the time domain. We derive a bound which gives insight into the frequency resolution or frequency sampling grid density problem, and also provides an analytical basis for determining an estimate of the filter length. Problems encountered with efficient transmitter implementation are discussed along with methods found useful in…

Probabilistic Distance for Mixtures of Independent Component Analyzers

Gonzalo Safont, Addisson Salazar, Luis Vergara +2 more

Signal Processing · IEEE Transactions on Neural Networks and Learning Systems · · cited 31× · DOI ↗ · more like this ·

Independent component analysis (ICA) is a blind source separation technique where data are modeled as linear combinations of several independent non-Gaussian sources. The independence and linear restrictions are relaxed using several ICA mixture models (ICAMMs) obtaining a two-layer artificial neural network structure. This allows for dependence between sources of different classes, and thus, a myriad of multidimensional probability density functions can be accurate modeled. This paper proposes a new probabilistic distance (PDI) between the parameters learned for two ICAMMs. The PDI is computed explicitly, unlike the popular Kullback-Leibler divergence (KLD) and other similar metrics, removing the need for numerical integration. Furthermore, the PDI is symmetric and bounded within 0 and 1, which enables its use as a posterior probability in fusion approaches. In this paper, the PDI is…

ICA Mixtures Applied to Ultrasonic Nondestructive Classification of Archaeological Ceramics

Addisson Salazar, Luis Vergara

Signal Processing · EURASIP Journal on Advances in Signal Processing · · cited 29× · DOI ↗ · more like this ·

We consider a classifier based on Independent Component Analysis Mixture Modelling (ICAMM) to model the feature joint-probability density. This classifier is applied to a challenging novel application: classification of archaeological ceramics. ICAMM gathers relevant characteristics that have general interest for material classification. It can deal with arbitrary forms of the underlying probability densities in the feature vector space as nonparametric methods can do. Mutual dependences among the features are modelled in a parametric form so that ICAMM can achieve good performance even with a training set of relatively small size, which is characteristic of parametric methods. Moreover, in the training stage, ICAMM can incorporate probabilistic semisupervision (PSS): labelling by an expert of a portion of the whole available training set of samples. These properties of ICAMM are…

A New Surrogating Algorithm by the Complex Graph Fourier Transform (CGFT)

Jordi Belda, Luis Vergara, Gonzalo Safont +2 more

Artificial Intelligence · Entropy · · cited 27× · DOI ↗ · more like this ·

The essential step of surrogating algorithms is phase randomizing the Fourier transform while preserving the original spectrum amplitude before computing the inverse Fourier transform. In this paper, we propose a new method which considers the graph Fourier transform. In this manner, much more flexibility is gained to define properties of the original graph signal which are to be preserved in the surrogates. The complex case is considered to allow unconstrained phase randomization in the transformed domain, hence we define a Hermitian Laplacian matrix that models the graph topology, whose eigenvectors form the basis of a complex graph Fourier transform. We have shown that the Hermitian Laplacian matrix may have negative eigenvalues. We also show in the paper that preserving the graph spectrum amplitude implies several invariances that can be controlled by the selected Hermitian…

Computing the Partial Correlation of ICA Models for Non-Gaussian Graph Signal Processing

Jordi Belda, Luis Vergara, Gonzalo Safont +1 more

Signal Processing · Entropy · · cited 27× · DOI ↗ · more like this ·

Conventional partial correlation coefficients (PCC) were extended to the non-Gaussian case, in particular to independent component analysis (ICA) models of the observed multivariate samples. Thus, the usual methods that define the pairwise connections of a graph from the precision matrix were correspondingly extended. The basic concept involved replacing the implicit linear estimation of conventional PCC with a nonlinear estimation (conditional mean) assuming ICA. Thus, it is better eliminated the correlation between a given pair of nodes induced by the rest of nodes, and hence the specific connectivity weights can be better estimated. Some synthetic and real data examples illustrate the approach in a graph signal processing context.

Graph Regularization Methods in Soft Detector Fusion

Addisson Salazar, Gonzalo Safont, Luis Vergara +1 more

Computer Networks and Communications · IEEE Access · · cited 26× · DOI ↗ · more like this ·

This paper presents a theoretical derivation of two new graph-based regularization methods for fusing the individual results of multiple detectors (two-class classifiers). The proposed approach considers linear combination of the individual detector statistics and its extension to a general nonlinear fusion method known as α-integration. A cost function that includes a mean-square error and a regularization term is minimized. The inclusion of the regularization term, which is based on graph signal processing, reduces the dispersion of the fused statistics, and thus improves the separation between the fused statistics corresponding to every detection hypothesis. The proposed methods (linear and non-linear regularized α-integration) are experimentally compared with commonly used classification methods (random forest, linear and quadratic discriminant analysis, and naive Bayes) and…

Signal processing on graphs for improving automatic credit card fraud detection

Luis Vergara, Addisson Salazar, Jordi Belda +3 more

Artificial Intelligence · · cited 26× · DOI ↗ · more like this ·

In this paper, several methods based on signal processing on graphs are proposed to improve the performance of credit card fraud detection. The proposed methods consist of a variant of the classic iterative amplitude adjusted Fourier transform (IAAFT) and two methods that we have called iterative surrogate signals on graph algorithms (ISSG). The objective is to generate surrogate samples from the original scarce fraud samples to improve the training of the detectors by lowering the variance of the estimate. A reliable augmentation of the target scarce population of frauds is important considering issues such as labeling cost; algorithm testing; data confidentiality; and constantly changing of patterns in the data streaming source. We have approached several scenarios with different legitimate and non-legitimate transaction ratios showing the feasibility of improving detection…

Combination of multiple detectors for credit card fraud detection

Addisson Salazar, Gonzalo Safont, Alberto Rodríguez +1 more

Artificial Intelligence · · cited 21× · DOI ↗ · more like this ·

This paper presents a signal processing framework for the problem of automatic credit card fraud detection. This is a critical problem affecting large financial companies that has increased due to the rapid expansion of information and communication technologies. The framework establishes relationships between signal processing and pattern recognition issues around a detection problem with a very low ratio between fraudulent and legitimate transactions. Solutions are proposed using fusion of scores which are related to the familiar likelihood ratio statistic. Moreover, the classical detection problem analyzed by receiving operating characteristic curves is mapped to real-world business requirements based on key performance indicators. A strong practical case which combines real and surrogate data is approached, including comparison of the proposed methods with standard methods.

Image Applications of Agglomerative Clustering Using Mixtures of Non-Gaussian Distributions

Addisson Salazar, Jorge Igual, Gonzalo Safont +2 more

Artificial Intelligence · · cited 20× · DOI ↗ · more like this ·

We present two applications in image processing of an agglomerative clustering method based on mixtures of non-Gaussian distributions. The method joins pair-wise the mixture models estimated for every cluster building a pyramidal or hierarchical structure by using the Kullback-Leibler divergence. This process can be related with the feedforward process of abstraction carried out by the brain. The applications consist of grouping images based on their content similarities and segmentation of regions of an image in similar areas. The capability of the method to distinguish between natural and artificial images is also demonstrated.

Learning Hierarchies from ICA Mixtures

Addisson Salazar, Jorge Igual, Luis Vergara +1 more

Signal Processing · IEEE International Conference on Neural Networks/IEEE ... International Conference on Neural Networks · · cited 19× · DOI ↗ · more like this ·

This paper presents a novel procedure to classify data from mixtures of independent component analyzers. The procedure includes two stages: learning the parameters of the mixtures (basis vectors and bias terms) and clustering the ICA mixtures following a bottom-up agglomerative scheme to construct a hierarchy for classification. The approach for the estimation of the source probability density function is non-parametric and the minimum kullback-Leibler distance is used as a criterion for merging clusters at each level of the hierarchy. Validation of the proposed method is presented from several simulations including ICA mixtures with uniform and Laplacian source distributions and processing real data from impact-echo testing experiments.

Blind detection of nonlinearities in multiple-echo ultrasonic signals

Ramón Miralles, Luis Vergara, Addisson Salazar +1 more

Signal Processing · IEEE Transactions on Ultrasonics Ferroelectrics and Frequency Control · · cited 8× · DOI ↗ · more like this ·

In this paper, we propose and analyze by means of simulations the use of surrogate data algorithms for blind detection of nonlinearities in multiple-echo ultrasonic signals. We assume a blind scheme so that no information about the input (emitted ultrasonic pulse) can be used. The metrics and equations that model some nonlinear situations are carefully reviewed. Also, closed form equations of the third-order metrics from a simplified second-order Volterra kernel are derived. Computer simulations show that the surrogate data technique is a potentially powerful tool for blind detection of nonlinearities in multiple-echo ultrasonic signals if adequate metrics are chosen. They also reveal interesting trade-offs among parameters that model ultrasonic systems and detection percentages.

Late Fusion for Improving Intrusion Detection in a Network Traffic Dataset

Addisson Salazar, Nancy Vargas, Gonzalo Safont +1 more

Computer Networks and Communications · 2021 International Conference on Computational Science and Computational Intelligence (CSCI) · · cited 5× · DOI ↗ · more like this ·

This paper presents a method for an intrusion detection system based on late fusion of classifiers. The proposed method was tested in the analysis of a network traffic dataset considering up to 14 classes of anomalous traffic. Several data quality issues were solved by preprocessing: missing data, non-numeric data types, and imbalance of the data classes. The high dimension of the data was reduced by a feature selection method. The results demonstrate the capabilities of late fusion to improve classification accuracy and stability of the intrusion traffic detection from the ones obtained by the individual classifiers.

Agglomerative clustering of defects in ultrasonic non-destructive testing using hierarchical mixtures of independent component analyzers

Addisson Salazar, Jorge Igual, Luis Vergara

Signal Processing · · cited 5× · DOI ↗ · more like this ·

This paper presents a novel procedure to classify materials with different defects, such as holes or cracks, from mixtures of independent component analyzers. The data correspond to the ultrasonic echo recorded after an impact by several sensors on the surface of the material. These signals are modelled by independent component analysis mixture models (ICAMM) for every kind of defect. After the ICAMM model is estimated for every defect, these are merged according to a distance measure that is obtained from the Kullback-Leibler divergence. The hierarchy obtained from the impact-echo data and the learning process allow different kinds of defective materials to be grouped consistently.

Mixtures of independent component analyzers for microarousal detection

Gonzalo Safont, Addisson Salazar, Luis Vergara +2 more

Signal Processing · IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI ...) · · cited 5× · DOI ↗ · more like this ·

This paper presents a study of the application of new variants of the Sequential Independent Analysis Mixture Models (SICAMM) to the modeling and classification of electroencephalographic (EEG) signals. The real application approached was the detection of microarousals in EEG signals from sleep apnea patients. In addition, the proposed methods were tested on synthetic data with probability density changing in time in order to imitate the intrinsic nonlinearity and nonstationarity of the EEG signals. Thus, sequential dependence and sensitivity analyses with controlled simulated data are included. The SICAMM-based methods were compared with Dynamic Bayesian Networks (DBN) implementing Gaussian mixture models and static versions of the methods. We demonstrate that the proposed methods obtain the best performance in microarousal detection and their parameters adapt better for EEG signal…

Constrained temporal extraction of the atrial rhythm in Atrial Fibrillation episodes

Raúl Llinares, Jorge Igual, Addisson Salazar +1 more

Signal Processing · PubMed · · cited 5× · DOI ↗ · more like this ·

The extraction of the Atrial Activity from the Ventricular Activity in Atrial Fibrillation episodes is a must for clinical analysis. We follow the semi Blind Source Extraction S-BSE approach to solve the problem. The proposed algorithm modifies the BSE contrast function to satisfy the prior knowledge about the spectral content of the atrial signal. The introduction of this prior allows obtaining a new algorithm with the following advantages: it allows the extraction of only the atrial component and it improves the quality of the recovered atrial signal in terms of spectral concentration as we show in the results.

Fusion of Visual and Textual Features for Table Header Detection in Handwritten Text Images

Addisson Salazar, José Ramón Fernández Prieto, Enrique Vidal +2 more

Computer Vision and Pattern Recognition · 2021 International Conference on Computational Science and Computational Intelligence (CSCI) · · cited 4× · DOI ↗ · more like this ·

This paper introduces a new procedure to improve table header detection in handwritten text images from the fusion of the posterior probabilities provided by two baseline classifiers. Each classifier considers a different modality, namely visual or textual features. Both baseline classifiers implements convolutional neural networks, particularly adopting the U-Net architecture. Four fusion methods are considered: the mean; linear discriminant analysis and random forest as meta-classifiers; and a recently developed method called alpha integration. The testing dataset consisted of 89 page images drawn from the Passau dataset. The improved performance provided by the fusion methods in the specific experiments is interesting considering the complexity of the challenging problem approached. In terms of area under the receiver operating characteristic curve the best results were obtained by…

New Applications of Late Fusion Methods for EEG Signal Processing

Gonzalo Safont, Addisson Salazar, Luis Vergara

Signal Processing · RiuNet (Politechnical University of Valencia) · · cited 4× · DOI ↗ · more like this ·

Decision fusion consists in the combination of the outputs of multiple classifiers into a common decision that is more precise or stable. In most cases, however, only classical fusion techniques are considered. This work compares the performance of several state-of-the-art fusion methods on new applications of automatic stage classification of several neuropsychological tests. The tests were staged into three classes: stimulus display, retention interval, and subject response. The considered late fusion methods were: alpha integration; copulas; Dempster-Shafer combination; independent component analysis mixture models; and behavior knowledge space. Late fusion was able to improve the performance for the task, with alpha integration yielding the most stable result.

A New Method for Fraud Detection in Credit Cards Based on Transaction Dynamics in Subspaces

Addisson Salazar, Gonzalo Safont, Luis Vergara

Artificial Intelligence · RiuNet (Politechnical University of Valencia) · · cited 4× · DOI ↗ · more like this ·

This paper presents a new method for fraud detection in credit cards based on exploiting the dynamics of the card transactions. We hypothesize different behavior models in the use of the card between legitimate clients and fraudsters that are registered in the sequential pattern that follows the transactions. The method considers analyses in subspaces defined by two or three variables recorded in the transactions. From these subspaces, several dynamic features, such as transaction velocity and acceleration, are estimated as input vectors for a classification process. Linear and quadratic discriminant analysis and random forest are implemented as single classifiers. All the single classification results obtained for each of the subspaces are late fused to obtain an overall result using alpha integration algorithm. The proposed method was evaluated using a subset of real data with a very…

New Perspectives of Pattern Recognition for Automatic Credit Card Fraud Detection

Addisson Salazar, Gonzalo Safont, Alberto Rodríguez +1 more

Artificial Intelligence · IGI Global eBooks · · cited 4× · DOI ↗ · more like this ·

Automatic credit card fraud detection (ACCFD) is a challenge issue that has been increasingly studied considering expanded potential of new technologies to emulate legitimate operations. Solution has to handle with fraud behavior changing in time; detection in data with very small fraud/legitimate operations ratio; and accomplish operation requirements of very low false alarm in real-time processing. In this chapter, main issues related with the problem of ACCFD and proposed solutions are discussed from theoretical and practical standpoints. The perspective of detection analyses from receiving operating characteristic curves and business key performance indicators are jointly analyzed. A new conceptual framework for ACCFD considering decision fusion and surrogate data is outlined including a case of study with different proportions of real and surrogate data. In addition, the…