We present a novel deep generative model based on non i.i.d. variational autoencoders that captures global dependencies among observations in a fully unsupervised fashion. In contrast to the recent semi-supervised alternatives for global modeling in deep generative models, our approach combines a mixture model in the local or data-dependent space and a global Gaussian latent variable, which lead us to obtain three particular insights. First, the induced latent global space captures interpretable disentangled representations with no user-defined regularization in the evidence lower bound (as in β-VAE and its generalizations). Second, we show that the model performs domain alignment to find correlations and interpolate between different databases. Finally, we study the ability of the global space to discriminate between groups of observations with non-trivial underlying structures, such as face images with shared attributes or defined sequences of digits images.

In this work, we address the problem of change-point detection (CPD) on high-dimensional, multi-source, and heterogeneous sequential data with missing values. We present a new CPD methodology based on local latent variable models and adaptive factorizations that enhances the fusion of multi-source observations with different statistical data-type and face the problem of high dimensionality. Our motivation comes from behavioral change detection in healthcare measured by smartphone monitored data and Electronic Health Records. Due to the high dimension of the observations and the differences in the relevance of each source information, other works fail in obtaining reliable estimates of the change-points location. This leads to methods that are not sensitive enough when dealing with interspersed changes of different intensity within the same sequence or partial missing components. Through the definition of local observation models (LOMs), we transfer the local CP information to homogeneous latent spaces and propose several factorizations that weight the contribution of each source to the global CPD. With the presented methods we demonstrate a reduction in both the detection delay and the number of not-detected CPs, together with robustness against the presence of missing values on a synthetic dataset. We illustrate its application on real-world data from a smartphone-based monitored study and add explainability on the degree of each source contributing to the detection.

Medical data sets are usually corrupted by noise and missing data. These missing patterns are commonly assumed to be completely random, but in medical scenarios, the reality is that these patterns occur in bursts due to sensors that are off for some time or data collected in a misaligned uneven fashion, among other causes. This paper proposes to model medical data records with heterogeneous data types and bursty missing data using sequential variational autoencoders (VAEs). In particular, we propose a new methodology, the Shi-VAE, which extends the capabilities of VAEs to sequential streams of data with missing observations. We compare our model against state-of-the-art solutions in an intensive care unit database (ICU) and a dataset of passive human monitoring. Furthermore, we find that standard error metrics such as RMSE are not conclusive enough to assess temporal models and include in our analysis the cross-correlation between the ground truth and the imputed signal. We show that Shi-VAE achieves the best performance in terms of using both metrics, with lower computational complexity than the GP-VAE model, which is the state-of-the-art method for medical records.

We present a framework for transfer learning based on modular variational Gaussian processes (GP). We develop a module-based method that having a dictionary of well fitted GPs, each model being characterised by its hyperparameters, pseudo-inputs and their corresponding posterior densities, one could build ensemble GP models without revisiting any data. Our method avoids undesired data centralisation, reduces rising computational costs and allows the transfer of learned uncertainty metrics after training. We exploit the augmentation of high-dimensional integral operators based on the Kullback-Leibler divergence between stochastic processes to introduce an efficient lower bound under all the sparse variational GPs, with different complexity and even likelihood distribution. The method is also valid for multi-output GPs, learning correlations a posteriori between independent modules. Extensive results illustrate the usability of our framework in large-scale and multi-task experiments, also compared with the exact inference methods in the literature.

This paper addresses the problem of change-point detection in sequences of high-dimensional and heterogeneous observations, which also possess a periodic temporal structure . Due to the dimensionality problem, when the time between change points is of the order of the dimension of the model parameters, drifts in the underlying distribution can be misidentified as changes. To overcome this limitation, we assume that the observations lie in a lower-dimensional manifold that admits a latent variable representation. In particular, we propose a hierarchical model that is computationally feasible, widely applicable to heterogeneous data and robust to missing instances. Additionally, the observations’ periodic dependencies are captured by non-stationary periodic covariance functions . The proposed technique is particularly well suited to (and motivated by) the problem of detecting changes in human behavior using smartphones and its application to relapse detection in psychiatric patients. Finally, we validate the technique on synthetic examples and we demonstrate its utility in the detection of behavioral changes using real data acquired by smartphones.

Change-point detection (CPD) aims to locate abrupt transitions in the generative model of a sequence of observations. When Bayesian methods are considered, the standard practice is to infer the posterior distribution of the change-point locations. However, for complex models (high-dimensional or heterogeneous), it is not possible to perform reliable detection. To circumvent this problem, we propose to use a hierarchical model, which yields observations that belong to a lower-dimensional manifold. Concretely, we consider a latent-class model with an unbounded number of categories, which is based on the chinese-restaurant process (CRP). For this model we derive a continual learning mechanism that is based on the sequential construction of the CRP and the expectation-maximization (EM) algorithm with a stochastic maximization step. Our numerical results show that the proposed method is able to recursively infer the number of underlying latent classes and perform CPD in a reliable manner.

This paper introduces a novel method to assess the social activity maintained by psychiatric patients using information and communication technologies. In particular, we model the daily usage patterns of phone calls and social and communication apps using point processes. We propose a novel nonhomogeneous Poisson process model with periodic (circadian) intensity function using a truncated Fourier series expansion, which is inferred using a trust-region algorithm. We also extend the model using a mixture of periodic intensity functions to cope with the different daily patterns of a person. The analysis of the usage of phone calls and social and communication apps of a cohort of 259 patients reveals common patterns shared among patients with relatively high homogeneity and differences among patient pathologies.

Multichannel intracavitary electrograms (EGMs) are acquired at the electrophysiology laboratory to guide radio frequency catheter ablation of patients suffering from atrial fibrillation. These EGMs are used by cardiologists to determine candidate areas for ablation (e.g., areas corresponding to high dominant frequencies or complex fractionated electrograms). In this paper, we introduce two hierarchical algorithms to retrieve the causal interactions among these multiple EGMs. Both algorithms are based on Granger causality, but other causality measures can be easily incorporated. In both cases, they start by selecting a root node, but they differ on the way in which they explore the set of signals to determine their cause-effect relationships: either testing the full set of unexplored signals (GS-CaRe) or performing a local search only among the set of neighbor EGMs (LS-CaRe). The ensuing causal model provides important information about the propagation of the electrical signals inside the atria, uncovering wavefronts and activation patterns that can guide cardiologists towards candidate areas for catheter ablation. Numerical experiments, on both synthetic signals and annotated real-world signals, show the good performance of the two proposed approaches.

This paper presents a novel method for predicting suicidal ideation from electronic health records (EHR) and ecological momentary assessment (EMA) data using deep sequential models. Both EHR longitudinal data and EMA question forms are defined by asynchronous, variable length, randomly sampled data sequences. In our method, we model each of them with a recurrent neural network, and both sequences are aligned by concatenating the hidden state of each of them using temporal marks. Furthermore, we incorporate attention schemes to improve performance in long sequences and time-independent pre-trained schemes to cope with very short sequences. Using a database of 1023 patients, our experimental results show that the addition of EMA records boosts the system recall to predict the suicidal ideation diagnosis from 48.13% obtained exclusively from EHR-based state-of-the-art methods to 67.78%. Additionally, our method provides interpretability through the t-distributed stochastic neighbor embedding (t-SNE) representation of the latent space. Furthermore, the most relevant input features are identified and interpreted medically.

We present a novel extension of multi-output Gaussian processes for handling heterogeneous outputs. We assume that each output has its own likelihood function and use a vector-valued Gaussian process prior to jointly model the parameters in all likelihoods as latent functions. Our multi-output Gaussian process uses a covariance function with a linear model of coregionalisation form. Assuming conditional independence across the underlying latent functions together with an inducing variable framework, we are able to obtain tractable variational bounds amenable to stochastic variational inference. We illustrate the performance of the model on synthetic data and two real datasets: a human behavioral study and a demographic high-dimensional dataset.

Wearable sensors are increasingly taking part in daily activities, not only because of the recent society health concern, but also due to their relevance in the medical industry. In this paper, a galvanic skin response (GSR) extraction technique has been developed in order to interpret electrodermal activity (EDA) records, which can be useful both for ambulatory and health applications. The core of the proposed approach is a novel feature extraction scheme that is based on a nonnegative sparse deconvolution of the observed GSR signals. Unlike previous approaches, the resulting SparsEDA algorithm is fast (immediately extracting the skin conductance level and response), efficient (being able to work with any sampling rate and signal length), and highly interpretable (due to the sparsity of the extracted phasic component of the GSR). Results on real data from 100 different subjects confirm the good performance of the method, which has been released through a free web-based code repository.

Multi-channel intracardiac electrocardiograms (electrograms) are sequentially acquired, at the electrophysiology laboratory, in order to guide radio frequency catheter ablation during heart surgery performed on patients with sustained atrial fibrillation (AF). These electrograms are used by cardiologists to determine candidate areas for ablation (e.g., areas corresponding to high dominant frequencies or complex fractionated electrograms). In this paper, we introduce a novel hierarchical algorithm for causality discovery among these multi-output sequentially acquired electrograms. The causal model obtained provides important information about the propagation of the electrical signals inside the heart, uncovering wavefronts and activation patterns that will serve to increase our knowledge about AF and guide cardiologists towards candidate areas for catheter ablation. Numerical results on synthetic signals, generated using the FitzHugh-Nagumo model, show the good performance of the proposed approach.

We consider the problem of daily human activity recognition (HAR) using multiple wireless inertial sensors, and specifically, HAR systems with a very low number of sensors, each one providing an estimation of the performed activities. We propose new Bayesian models to combine the output of the sensors. The models are based on a soft outputs combination of individual classifiers to deal with the small number of sensors. We also incorporate the dynamic nature of human activities as a first-order homogeneous Markov chain. We develop both inductive and transductive inference methods for each model to be employed in supervised and semisupervised situations, respectively. Using different real HAR databases, we compare our classifiers combination models against a single classifier that employs all the signals from the sensors. Our models exhibit consistently a reduction of the error rate and an increase of robustness against sensor failures. Our models also outperform other classifiers combination models that do not consider soft outputs and an Markovian structure of the human activities.

Monte Carlo (MC) methods are useful tools for Bayesian inference and stochastic optimization that have been widely applied in signal processing and machine learning. A well-known class of MC methods are Markov Chain Monte Carlo (MCMC) algorithms. In this work, we introduce a novel parallel interacting MCMC scheme, where the parallel chains share information, thus yielding a faster exploration of the state space. The interaction is carried out generating a dynamic repulsion among the “smelly” parallel chains that takes into account the entire population of current states. The ergodicity of the scheme and its relationship with other sampling methods are discussed. Numerical results show the advantages of the proposed approach in terms of mean square error, robustness w.r.t. to initial values and parameter choice.

Hidden Markov Models (HMMs) are one of the most important techniques to model and classify sequential data. Maximum Likelihood (ML) and (parametric and non-parametric) Bayesian estimation of the HMM parameters suffers from local maxima and in massive datasets they can be specially time consuming. In this paper, we extend the spectral learning of HMMs, a moment matching learning technique free from local maxima, to discriminative HMMs. The resulting method provides the posterior probabilities of the classes without explicitly determining the HMM parameters, and is able to deal with missing labels. We apply the method to Human Activity Recognition (HAR) using two different types of sensors: portable inertial sensors, and fixed, wireless binary sensor networks. Our algorithm outperforms the standard discriminative HMM learning in both complexity and accuracy.

Automated screening systems are commonly used to detect some agent in a sample and take a global decision about the subject (e.g., ill/healthy) based on these detections. We propose a Bayesian methodology for taking decisions in (sequential) screening systems that considers the false alarm rate of the detector. Our approach assesses the quality of its decisions and provides lower bounds on the achievable performance of the screening system from the training data. In addition, we develop a complete screening system for sputum smears in tuberculosis diagnosis, and show, using a real-world database, the advantages of the proposed framework when compared to the commonly used count detections and threshold approach.

Negative co-occurrence is a common phenomenon in many signal processing applications. In some cases the signals involved are sparse, and this information can be exploited to recover them. In this paper, we present a sparse learning approach that explicitly takes into account negative co-occurrence. This is achieved by adding a novel penalty term to the LASSO cost function based on the cross-products between the reconstruction coefficients. Although the resulting optimization problem is non-convex, we develop a new and efficient method for solving it based on successive convex approximations. Results on synthetic data, for both complete and overcomplete dictionaries, are provided to validate the proposed approach.

The reduction of the energy consumption in the domain of the embedded systems is becoming the most important design goal due to the increasing use of battery powered consumer devices. Previous research has pointed out the instruction memory organisation as one of the major sources of energy consumption of the embedded systems. Due to this fact, the introduction of any enhancement in this component of the system becomes crucial in order to decrease this energy bottleneck. The purpose of this paper is to present a highlevel energy analysis of the loop buffer schemes that exist in the embedded systems. The crucial energy analysis that is presented in this paper not only proposes a method to evaluate different loop buffer schemes for a certain application, but also guides embedded systems designers to make the correct decision in the trade-offs that exist between the energy budget, the required performance, and the area cost of the embedded system. Experimental results used in this analysis show that, the search of energy savings (up to 76%) has to take into account the performance penalty, the area cost, and the impact of the implementation technology in order to choose the most suitable enhancement that has to be introduced in the instruction memory organisation from the point of view of the energy consumption.

Abstract—In this paper, we provide a unified study of the application of kernel density estimators to supervised linear feature extraction by means of criteria inspired by information and detection theory. We enrich this study by the incorporation of two novel criteria to the study: the mutual information and the likelihood ratio test, and perform both a theoretical and a experimental comparison between the new methods and other ones previously described in the literature. The impact of the bandwidth selection of the density estimator in the classification performance is discussed. Some theoretical results that bound classification performance as a function or mutual information are also compiled. A set of experiments on different realworld datasets allow us to perform an empirical comparison of the methods, in terms of both accuracy and computational complexity. We show the suitability of these methods to determine the dimension of the sub-space that contains the discriminative information.

Instruction memory organization is pointed out as one of the major sources of energy consumption in embedded systems. As embedded systems are characterized by restrictive resources and low energy budget, any enhancement in this component allows not only to decrease the total energy consumption, but also to have a better distribution of the energy budget throughout the system. This paper presents a self-tuning banked loop buffer architecture, which is based on a run-time loop buffer controller that optimizes both the dynamic and leakage energy consumption of the instruction memory organization. Results show that using banking in loop buffer architectures leads to higher reduction in the total energy consumption of the instruction memory organization if the tuning approach is applied sparingly. Based on post-layout simulations, our approach improves the total energy consumption by average of 20% in comparison with a loop buffer architecture based on a single monolithic memory, and more than 90% in comparison with instruction memory organizations without loop buffer architectures.

In some applications, the probability of error of a given classifier is too high for its practical application, but we are allowed to gather more independent test samples from the same class to reduce the probability of error of the final decision. From the point of view of hypothesis testing, the solution is given by the Neyman-Pearson lemma. However, there is no equivalent result to the Neyman-Pearson lemma when the likelihoods are unknown, and we are given a training dataset. In this brief, we explore two alternatives. First, we combine the soft (probabilistic) outputs of a given classifier to produce a consensus labeling for K test samples. In the second approach, we build a new classifier that directly computes the label for K test samples. For this second approach, we need to define an extended input space training set and incorporate the known symmetries in the classifier. This latter approach gives more accurate results, as it only requires an accurate classification boundary, while the former needs an accurate posterior probability estimate for the whole input space. We illustrate our results with well-known databases.

We implement a fixed-point real-time identification system and provide tools for the optimal design of exponential lookup-tables. This intelligent surveillance system is based on infrared image processing, which allows to detect and track people and trigger different actions depending on the region of the monitored area in which they are. The system automatically segments the body to get the face and includes a face classifier based on the support vector machine.

We consider a large-scale, dense sensor network with two types of binary sensors with different discrimination performance and costs. We want to analyze what is the optimal proportion of both sensors in a target detection problem by assigning a given total cost per measurement. We obtain the hypothesis test, and we obtain the test performance using large deviation bounds. We found that the optimal proportion is "winner takes all" like. The sensor with the best performance/cost ratio is selected.

This paper proposes a new cost function for supervised training of neural networks in binary classification applications. This cost function aims at reducing the probability of classification error by reducing the overlap between distributions of the soft output for each class. The non parametric Parzen window method, with Gaussian kernels, is used to estimate the distributions from the training data set. The cost function has been implemented in a GRBF neural network and has been tested in a motion detection application from low resolution infrared images, showing some advantages with respect to the conventional mean squared error cost function and also with respect to the support vector machine, a reference binary classifier.

In this paper, we present a novel scheme for linear feature extraction in classification. The method is based on the maximization of the mutual information (MI) between the features extracted and the classes. The sum of the MI corresponding to each of the features is taken as an heuristic that approximates the MI of the whole output vector. Then, a component-by-component gradient-ascent method is proposed for the maximization of the MI, similar to the gradient-based entropy optimization used in independent component analysis (ICA). The simulation results show that not only is the method competitive when compared to existing supervised feature extraction methods in all cases studied, but it also remarkably outperform them when the data are characterized by strongly nonlinear boundaries between classes.

This paper proposes a twofold approach for therapeutic drug monitoring (TDM) of kidney recipients using support vector machines (SVMs), for both predicting and detecting Cyclosporine A (CyA) blood concentrations. The final goal is to build useful, robust, and ultimately understandable models for individualizing the dosage of CyA. We compare SVMs with several neural network models, such as the multilayer perceptron (MLP), the Elman recurrent network, finite/infinite impulse response networks, and neural network ARMAX approaches. In addition, we present a profile-dependent SVM (PD-SVM), which incorporates a priori knowledge in both tasks. Models are compared numerically, statistically, and in the presence of additive noise. Data from 57 renal allograft recipients were used to develop the models. Patients followed a standard triple therapy, and CyA trough concentration was the dependent variable. The best results for the CyA blood concentration prediction were obtained using the PD-SVM (mean error of 0.36 ng/mL and root-mean-square error of 52.01 ng/mL in the validation set) and appeared to be more robust in the presence of additive noise. The proposed PD-SVM improved results from the standard SVM and MLP, specially significant (both numerical and statistically) in the one-against-all scheme. Finally, some clinical conclusions were obtained from sensitivity rankings of the models and distribution of support vectors. We conclude that the PD-SVM approach produces more accurate and robust models than do neural networks. Finally, a software tool for aiding medical decision-making including the prediction models is presented

In this paper, we apply the methodology of cross-validation maximum likelihood estimation to the problem of multivariate kernel density modeling. We provide a fixed point algorithm to find the covariance matrix for a Gaussian kernel according to this criterion. We show that the algorithm leads to accurate models in terms of entropy estimation and Parzen classification. By means of a set of experiments, we show that the method considerably improves the performance traditionally expected from Parzen classifiers. The accuracy obtained in entropy estimation suggests its usefulness in ICA and other information-theoretic signal processing techniques

We address the problem of tracking a maneuvering target that moves along a region monitored by a sensor network, whose nodes (including both the sensors and any data fusion centers, DFCs) are located at unknown positions. Thus, the node locations and the target track must be estimated jointly without the aid of beacons. We assume that the network consists of a collection of sensors and at least four DFCs. Each DFC collects and integrates the sensor measurements and can exchange data with the other DFCs. Within this setup, we propose a three-stage Monte Carlo method to (i) acquire rough initial estimates of the network node locations, (ii) track the target and refine the node position estimates individually at each DFC and (iii) fuse the results obtained by all the DFCs. The validity of the method is illustrated by computer simulations of a network of power-aware sensors and exactly four DFCs.

In this paper we propose a novel approach to distributed detection using learning-based local classifiers and likelihood ratio test (LRT) based fusion center. Local detector's soft outputs are not restricted to have any probabilistic meaning, so even pure discriminative training method can be used. We propose to estimate the conditional densities of the soft output of the local classifiers to formulate the LRT in the fusion center. Also, we suggest simple censoring schemes that take into account the learning-based approaches problem of the slow convergence of tails of learned distributions. The Neyman Pearson (NP) and the sequential probability radio tests are developed for this approach and NP performance is analyzed. The generality of the proposed procedure is illustrated in an example outside the typical field of sensor networks: the automated infectious tuberculosis (TB) diagnosis using local detections of TB bacilli in microscopic images

In this work, we consider the problem of detection in a sensor network with censored transmission. The motivation for this is the improvement of the energy efficiency by means of allowing just positive detection transmission. Both Bayes and NP (Neyman-Pearson) tests are developed and the performance of the NP test is studied using large deviation bounds on the error probability. We also show that these bounds using the KL divergence between the probability density functions of the observations under each hypothesis might be used as a criteria for determining the optimal exploration area or even the optimal strategy for energy efficiency. The "Spanish hat" model for the probability of detection is used to exemplify the performance of the proposed bound.

An iterative reweighted least squares (IRWLS) procedure recently proposed is shown to converge to the support vector machine solution. The convergence to a stationary point is ensured by modifying the original IRWLS procedure.

In this paper, a new method for the simultaneous learning of a function and its derivative is presented. The method, setting out the problem inside of the Support Vector Machine (SVM) framework, relies on the kernel-based Support Vector expansion. The resultant optimization problem is solved by a computationally efficient Iterative Re-Weighted Least Squares (IRWLS) algorithm.

We consider the problem of binary distributed detection in the context of large-scale, dense sensor networks. We propose to model the probability of detection in each sensor, p/sub d/, as a function of the distance between the sensor and the source or target to be detected. We derive the Bayesian fusion rule under that model. We also derive, using the asymptotic Gaussianity of the log-likelihood ratio, the Neyman-Pearson fusion rule. The performance of both tests is analyzed using large deviation bounds on the error probability and a parametric approximation to p/sub d/. The main conclusions of the analysis of these bounds are that, for designing efficient tests in terms of energy consumption, (1) the sensors must be grouped in areas of the order of the range of the local detectors, and, (2) the sensor must be configured to achieve the best local discrimination between hypothesis, independently of the configuration of the network.

We present the convergence demonstration of the iterative re-weighted least squares (IRWLS) procedure to the SVM solution, to propose two modifications, which significantly reduces the runtime complexity of the IRWLS. We show by means of computer experiments that the convergence can be speed up between two and eight times compare to the standard IRWLS procedure.

An adaptive, convex linear combination of the outputs of a standard least mean square (LMS) equalizer and a sigmoidal equalizer is proposed. This procedure results in improving the speed of the LMS equalizer while retaining the low steady-state error of the sigmoidal filter. Appropriate adaption schemes for both of the filters and for the combination parameters are established. Simulations of practical communication applications demonstrate the effectiveness of this adaptive combination.

T IS universally recognized that learning is an essential process both for living beings and for physical machines or computational programs to be efficient in carrying out complex tasks or when working in changing environments: Consequently, more and more intensive research is being dedicated to its many aspects. Most of this research adopts a particular point of view: biological, psychological, educational, mathematical, algorithmic, computational, etc. Nevertheless, many researchers consider that perspectives coming from other scientific fields are important, sometimes critical, to produce significant advances and useful results. The conference Learning’02 was intended to provide a forum for interdisciplinary study and discussion of the different aspects of learning: during two-and-a-half days, it hosted invited lectures by well-known experts, poster sessions, and round tables to discuss how the different approaches can help one another. This special issue contains seven papers originally presented at Learning’02, selected by members of the committees of the conference, and have all been extended by their authors and revised following the standard procedures for peer reviewing in IEEE journals.

We tackle the problem of detecting sources of combustion in high definition multispectral medium wavelength infrared (MWIR) (3-5 /spl mu/m) images. We present a novel approach to this problem consisting of processing the images block-wise using a new technique that we call supervised principal component analysis (SPCA) to get the components of these blocks. This outperforms state-of-the-art methods with a significant reduction in the complexity of the whole scheme. As a classifier, we propose the use of a support vector machine (SVM) comparing the results from both its novelty-detection and binary non-linear versions. High performance is achieved from a small set of components.

We present a new approach to nonparametric spectral estimation on the basis of the support vector method (SVM). A reweighted least squares error formulation avoids the computational limitations of quadratic programming. The application to a synthetic example and to a digital communication problem shows the robustness of the SVM spectral analysis algorithm.

In this paper, we propose a general technique for solving support vector classifiers (SVCs) for an arbitrary loss function, relying on the application of an iterative reweighted least squares (IRWLS) procedure. We further show that three properties of the SVC solution can be written as conditions over the loss function. This technique allows the implementation of the empirical risk minimization (ERM) inductive principle on large margin classifiers obtaining, at the same time, very compact (in terms of number of support vectors) solutions. The improvements obtained by changing the SVC loss function are illustrated with synthetic and real data examples.

We present an approach to solve the v-SVR. It is based on an iterative re-weighted least squares (IRWLS) procedure, which is simple to implement and can be tuned to the usual /spl nu/-SVR solution. The IRWLS procedure is much more efficient (computational load) than quadratic programming techniques, which are usually employed to solve it.

The subject of this paper is the compensation for nonlinearities in digital communication systems by means of predistortion. In this work, we apply the generalized cerebellar model articulation controller (GCMAC) to simplify and accelerate the predistorter convergence. The range of analyzed predistorters includes: 1) a symbol-rate data predistorter that, for a given time span, achieves a similar level of compensation provided by present techniques, but with faster convergence; 2) a fractionally spaced data predistorter that controls, at the same time, the signal constellation and the transmitted spectrum; 3) a decision-feedback scheme that compensates for remote nonlinearities; and 4) a digital signal data predistorter. The performance of the proposed data and signal predistorters is evaluated using typical linear and nonlinear modulated transmitted signals such as QAM and GMSK.

An iterative block training method for support vector classifiers (SVCs) based on weighted least squares (WLS) optimization is presented. The algorithm, which minimizes structural risk in the primal space, is applicable to both linear and nonlinear machines. In some nonlinear cases, it is necessary to previously find a projection of data onto an intermediate-dimensional space by means of either principal component analysis or clustering techniques. The proposed approach yields very compact machines, the complexity reduction with respect to the SVC solution is especially notable in problems with highly overlapped classes. Furthermore, the formulation in terms of WLS minimization makes the development of adaptive SVCs straightforward, opening up new fields of application for this type of model, mainly online processing of large amounts of (static/stationary) data, as well as online update in nonstationary scenarios (adaptive solutions). The performance of this new type of algorithm is analyzed by means of several simulations.

The support vector classifier is a new tool to solve classification problems, giving the classification boundary as a linear combination of the training samples. In non-separable problems with highly overlapped classes, the achieved classifiers are oversized. In this paper, we proposed to change the support vector classifier penalty function by an hyperbolic tangent one, obtaining as a result of the training phase a reduced support vector classifier with the same performance as the original one.

In this communication we present a new algorithm for solving Support Vector Classifiers (SVC) with large training data sets. The new algorithm is based on an Iterative Re-Weighted Least Squares procedure which is used to optimize the SVc. Moreover, a novel sample selection strategy for the working set is presented, which randomly chooses the working set among the training samples that do not fulfill the stopping criteria. The validity of both proposals, the optimization procedure and sample selection strategy, is shown by means of computer experiments using well-known data sets.

Based on a Gaussian mixture model for the reflectivity sequence, the authors present a new technique for blind deconvolution of seismic data. The method obtains a deconvolution filter that maximizes at its output a measure of the relative entropy between the proposed Gaussian mixture and a pure Gaussian distribution. A new updating procedure for the mixture parameters is included in the algorithm: it allows one to apply the algorithm without any prior knowledge about the signal and noise. A simulation example illustrates the performance of the proposed method.

This paper deals with the compensation for the nonlinear distortion introduced by power-efficient amplifiers on linear modulation by means of equalization. We propose a new equalizer based on a reduced-complexity network called the generalized cerebellar model arithmetic computer (GMAC) GCMAC. The GCMAC-based equalizer is compared with other well-known structures such as the Volterra filter and the multi-layer perceptron. Extensive computer simulations have been carried out. The obtained results show the effectiveness of the proposed structure to compensate for strong nonlinearities.

The cerebellar model articulation controller (CMAC) is a simple and fast neural-network based on local approximations. However, its rigid structure reduces its accuracy of approximation and speed of convergence with heterogeneous inputs. In this paper, we propose a generalized CMAC (GCMAC) network that considers different degrees of generalization for each input. Its representation abilities are analyzed, and a set of local relationships that the output function must satisfy are derived. An adaptive growing method of the network is also presented. The validity of our approach and methods are shown by some simulated examples.

This paper presents a new Bussgang-type technique for blind deconvolution of spiky signals. Based on a Gaussian mixture model for the spiky signal, the method obtains a deconvolution filter and a zero-memory nonlinearity to estimate the signal. A new updating procedure for the mixture parameters (and, therefore, for the nonlinear estimator) is included in the algorithm: it allows to apply the algorithm without any prior knowledge about the signal and noise. A simulation example illustrates the performance of the proposed method.

Sequence extrapolation is a classical problem in signal processing. The application of the expectation-maximization (EM) algorithm to estimate the signal autocorrelation in Wiener extrapolation allows us to view previous techniques as particular cases, to modify them, and to introduce new ones offering better performance with a moderate increase in computational load.

The authors explore the possibility of introducing a novel family of algorithms for data blind equalization based on minimizing phase error cost functions. A simple discussion of the difficulties involved and how to avoid them serves to introduce the family. Then, the authors discuss the general and particular characteristics of these schemes, and show the performance of two representative particular cases by means of simulation examples. Finally, they propose some basic ideas on reducing the limitation of these phase constraining algorithms.>

The drawbacks of using the adaptive delay filter for plant identification are discussed. These drawbacks are due to regression and gain estimation errors and the filter's limitations under colored excitation hypotheses. After proposing simple solutions for both drawbacks, the authors show how the limitations can be solved by using different known techniques, such as dynamic programming or by applying adaptively thresholded gradient-type algorithms. The direct use of the latter algorithms is also presented as an alternative. Some simulation results illustrate the behaviour of the proposed modifications and extensions.>