Search

Data fusion refers to the joint analysis of multiple datasets that provide different (e.g., complementary) views of the same task. In general, it can extract more information than separate analyses can. Jointly analyzing electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) measurements has been proved to be highly beneficial to the study of the brain function, mainly because these neuroimaging modalities have complementary spatiotemporal resolution: EEG offers good temporal resolution while fMRI is better in its spatial resolution. The EEG–fMRI fusion methods that have been reported so far ignore the underlying multiway nature of the data in at least one of the modalities and/or rely on very strong assumptions concerning the relation of the respective datasets. For example, in multisubject analysis, it is commonly assumed that the hemodynamic response function is a priori known for all subjects and/or the coupling across corresponding modes is assumed to be exact (hard). In this article, these two limitations are overcome by adopting tensor models for both modalities and by following soft and flexible coupling approaches to implement the multimodal fusion. The obtained results are compared against those of parallel independent component analysis and hard coupling alternatives, with both synthetic and real data (epilepsy and visual oddball paradigm). Our results demonstrate the clear advantage of using soft and flexible coupled tensor decompositions in scenarios that do not conform with the hard coupling assumption. © 2021 The Authors. Human Brain Mapping published by Wiley Periodicals LLC.

Tensor rank learning for canonical polyadic decomposition (CPD) has long been deemed as an essential yet challenging problem. In particular, since thetensor rank controls the complexity of the CPD model, its inaccurate learning would cause overfitting to noise or underfitting to the signal sources, and even destroy the interpretability of model parameters. However, the optimal determination of a tensor rank is known to be a non-deterministic polynomial-time hard (NP-hard) task. Rather than exhaustively searching for the best tensor rank via trial-and-error experiments, Bayesian inference under the Gaussian-gamma prior was introduced in the context of probabilistic CPD modeling, and it was shown to be an effective strategy for automatic tensor rank determination. This triggered flourishing research on other structured tensor CPDs with automatic tensor rank learning. On the other side of the coin, these research works also reveal that the Gaussian-gamma model does not perform well for high-rank tensors and/or low signal-to-noise ratios (SNRs). To overcome these drawbacks, in this paper, we introduce a more advanced generalized hyperbolic (GH) prior to the probabilistic CPD model, which not only includes the Gaussian-gamma model as a special case, but also is more flexible to adapt to different levels of sparsity. Based on this novel probabilistic model, an algorithm is developed under the framework of variational inference, where each update is obtained in a closed-form. Extensive numerical results, using synthetic data and real-world datasets, demonstrate the significantly improved performance of the proposed method in learning both low as well as high tensor ranks even for low SNR cases. © 1991-2012 IEEE.

This special issue covers promising developments in the related areas of machine learning (ML) and deep neural networks (NNs) and offers possible paths for the future. Deep neural networks (NNs) have offered performances much beyond of the incremental nature as compared to previous techniques. The scope of this issue is to put together a number of papers, written by world experts in the field, that try to provide some answers, to the possible extent, to the previously discussed problems, and at the same time to summarize research that has been carried in the related areas. The topics covered range from theory and algorithms to applications and hardware implementations.

In this paper, we introduce a novel methodology for the analysis of task-related fMRI data. In particular, we propose an alternative way for constructing the design matrix, based on the newly suggested Information-Assisted Dictionary Learning (IADL) method. This technique offers an enhanced potential, within the conventional GLM framework, (a) to efficiently cope with uncertainties in the modeling of the hemodynamic response function, (b) to accommodate unmodeled brain-induced sources, beyond the task-related ones, as well as potential interfering scanner-induced artifacts, uncorrected head-motion residuals and other unmodeled physiological signals, and (c) to integrate external knowledge regarding the natural sparsity of the brain activity that is associated with both the experimental design and brain atlases. The capabilities of the proposed methodology are evaluated via a realistic synthetic fMRI-like dataset, and demonstrated using a test case of a challenging fMRI study, which verifies that the proposed approach produces substantially more consistent results compared to the standard design matrix method. A toolbox extension for SPM is also provided, to facilitate the use and reproducibility of the proposed methodology. © 2021

Advances in computer and communications technology have deeply affected the way we communicate. Social media have emerged as a major means of human communication. However, a major limitation in such media is the lack of non-verbal stimuli, which sometimes hinders the understanding of the message, and in particular the associated emotional content. In an effort to compensate for this, people started to use emoticons, which are combinations of keyboard characters that resemble facial expressions, and more recently their evolution: emojis, namely, small colorful images that resemble faces, actions and daily life objects. This paper presents evidence of the effect of emojis on memory retrieval through a functional Magnetic Resonance Imaging (fMRI) study. A total number of fifteen healthy volunteers were recruited for the experiment, during which successive stimuli were presented, containing words with intense emotional content combined with emojis, either with congruent or incongruent emotional content. Volunteers were asked to recall a memory related to the stimulus. The study of the reaction times showed that emotional incongruity among word+emoji combinations led to longer reaction times in memory retrieval compared to congruent combinations. General Linear Model (GLM) and Blind Source Separation (BSS) methods have been tested in assessing the influence of the emojis on the process of memory retrieval. The analysis of the fMRI data showed that emotional incongruity among word +emoji combinations activated the Broca’s area (BA44 and BA45) in both hemispheres, the Supplementary Motor Area (SMA) and the inferior prefrontal cortex (BA47), compared to congruent combinations. Furthermore, compared to pseudowords, word+emoji combinations activated the left Broca’s area (BA44 and BA45), the amygdala, the right temporal pole (BA48) and several frontal regions including the SMA and the inferior prefrontal cortex. © 2020 Chatzichristos et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

The next generation of telecommunication networks will integrate the latest developments and emerging advancements in telecommunications connectivity infrastructures. In this article, we discuss the transformation and convergence of the fifth-generation (5G) mobile network and the internet of things technologies, toward the emergence of the smart sixth-generation (6G) networks which will employ AI to optimize and automate their operation. © 1999-2012 IEEE.

Gaussian processes (GPs) for machine learning have been studied systematically over the past two decades. However, kernel design for GPs and the associated hyper-parameters optimization are still difficult, and to a large extent open problems. We consider GP regression for time series modeling and analysis. The underlying stationary kernel is approximated closely by a new grid spectral mixture (GSM) kernel, which is a linear combination of low-rank sub-kernels. In the case where a large number of the involved sub-kernels are used, either the Nyström or the random Fourier feature approximations can be adopted to reduce the required computer storage. The unknown GP hyper-parameters consist of the nonnegative weights of all sub-kernels as well as the noise variance, and they are determined through the maximum-likelihood estimation method. Two optimization methods for solving the unknown hyper-parameters are introduced, including a sequential majorization-minimization (MM) method and a nonlinearly constrained alternating direction method of multipliers (ADMM). Experimental results, based on various time series datasets, corroborate that the proposed GSM kernel-based GP regression model outperforms several benchmarks in terms of prediction accuracy and numerical stability. © 1991-2012 IEEE.

Background: The growing interest in neuroimaging technologies generates a massive amount of biomedical data of high dimensionality. Tensor-based analysis of brain imaging data has been recognized as an effective analysis that exploits its inherent multi-way nature. In particular, the advantages of tensorial over matrix-based methods have previously been demonstrated in the context of functional magnetic resonance imaging (fMRI) source localization. However, such methods can also become ineffective in realistic challenging scenarios, involving, e.g., strong noise and/or significant overlap among the activated regions. Moreover, they commonly rely on the assumption of an underlying multilinear model generating the data. New method: This paper aims at investigating the possible gains from exploiting the 4-dimensional nature of the brain images, through a higher-order tensorization of the fMRI signal, and the use of less restrictive generative models. In this context, the higher-order block term decomposition (BTD) and the PARAFAC2 tensor models are considered for the first time in fMRI blind source separation. A novel PARAFAC2-like extension of BTD (BTD2) is also proposed, aiming at combining the effectiveness of BTD in handling strong instances of noise and the potential of PARAFAC2 to cope with datasets that do not follow the strict multilinear assumption. Comparison with existing methods: The methods were tested using both synthetic and real data and compared with state of the art methods. Conclusions: The simulation results demonstrate the effectiveness of BTD and BTD2 for challenging scenarios (presence of noise, spatial overlap among activation regions and inter-subject variability in the haemodynamic response function (HRF)). © 2018 Elsevier B.V.

Objectives: A special vocal ornament in Byzantine chant (BC), the single cycle ornamentation structure (SCOS), is defined and compared with the vibrato with respect to its time (rate, extent) and spectral (slope [SS], relative speaker's formant [SPF] level, formant frequencies [Fi] and bandwidths [Bi], and noise-to-harmonics ratio [NHR]) characteristics. Study Design: This is a comparative study between the vocal ornaments of SCOS and vibrato, of which time and spectral acoustic parameters were measured, statistically analyzed, and compared. Methods: From the same hymn recordings chanted by four chanters, the SS, SPF level, FFi, FBi, and NHR difference values between the vocal ornament and its neighbor steady note, and the rate and extent, were compared with those of vibrato. Results: The mean extent values for SCOS were found to be almost double the corresponding values for vibrato, and the rate of SCOS tends to be different from the rate of vibrato. The difference values of: 1) the NHR, 2) the spectral slope, and 3) the SPF level, between the vocal ornament and its neighbor steady note were found to be: 1) higher for SCOS, 2) mainly lower for SCOS, and 3) lower for SCOS, respectively. No significant differences were detected for the FFi and FBi. The FF1 differences tend to be negative in both ornaments indicating a formant tuning effect. Conclusions: A new vocal ornament (SCOS) in BC is studied, of which the extent, NHR (HNR), the spectral slope, and the SPF level are different compared to those of vibrato. © 2019 The Voice Foundation

In this paper, we propose a novel unsupervised pre-training method to learn the brain dynamics using a deep learning architecture named residual D-net. As it is often the case in medical research, in contrast to typical deep learning tasks, the size of the resting-state functional Magnetic Resonance Image (rs-fMRI) datasets for training is limited. Thus, the available data should be very efficiently used to learn the complex patterns underneath the brain connectivity dynamics. To address this issue, we use residual connections to alleviate the training complexity through recurrent multi-scale representation and pre-training the architecture unsupervised way. We conduct two classification tasks to differentiate early and late stage Mild Cognitive Impairment (MCI) from Normal healthy Control (NC) subjects. The experiments verify that our proposed residual D-net indeed learns the brain connectivity dynamics, leading to significantly higher classification accuracy compared to previously published techniques. © Springer Nature Switzerland AG 2019.

We present a novel diffusion scheme for online kernel-based learning over networks. So far, a major drawback of any online learning algorithm, operating in a reproducing kernel Hilbert space (RKHS), is the need for updating a growing number of parameters as time iterations evolve. Besides complexity, this leads to an increased need of communication resources in a distributed setting. In contrast, we propose to approximate the solution as a fixed-size vector (of larger dimension than the input space) using the previously introduced framework of random Fourier features. This paves the way to use standard linear combine-then-adapt techniques. To the best of our knowledge, this is the first time that a complete protocol for distributed online learning in RKHS is presented. Conditions for asymptotic convergence and boundness of the networkwise regret are also provided. The simulated tests illustrate the performance of the proposed scheme. © 1991-2012 IEEE.

We consider the task of robust nonlinear regression in the presence of both inlier noise and outliers. Assuming that the unknown nonlinear function belongs to a reproducing kernel Hilbert space, our goal is to estimate the set of the associated unknown parameters. Due to the presence of outliers, common techniques such as the kernel ridge regression (KRR) or the support vector regression turn out to be inadequate. Instead, we employ sparse modeling arguments to explicitly model and estimate the outliers, adopting a greedy approach. The proposed robust scheme, i.e., kernel greedy algorithm for robust denoising (KGARD), is inspired by the classical Orthogonal matching pursuit (OMP) algorithm. Specifically, the proposed method alternates between a KRR task and an OMP-like selection step. Theoretical results concerning the identification of the outliers are provided. Moreover, KGARD is compared against other cutting edge methods, where its performance is evaluated via a set of experiments with various types of noise. Finally, the proposed robust estimation framework is applied to the task of image denoising, and its enhanced performance in the presence of outliers is demonstrated. © 2017 IEEE.

The growing use of neuroimaging technologies generates a massive amount of biomedical data that exhibit high dimensionality. Tensor-based analysis of brain imaging data has been proved quite effective in exploiting their multiway nature. The advantages of tensorial methods over matrix-based approaches have also been demonstrated in the context of functional magnetic resonance imaging (fMRI) data analysis. However, such methods can become ineffective in demanding scenarios, involving, e.g., strong noise and/or significant overlapping of activated regions. This paper aims at investigating the possible gains that can be obtained from a better exploitation of the spatial dimension, through a higher (than 3)-order tensor modeling of the fMRI signals. In this context, a higher-order Block Term Decomposition (BTD) is applied, for the first time in fMRI analysis. Its effectiveness in handling strong instances of noise is demonstrated via extensive simulation results. © Springer International Publishing AG 2017.

This paper focuses on the development of novel greedy techniques for distributed learning under sparsity constraints. Greedy techniques have widely been used in centralized systems due to their low computational requirements and at the same time their relatively good performance in estimating sparse parameter vectors/signals. The paper reports two new algorithms in the context of sparsity-aware learning. In both cases, the goal is first to identify the support set of the unknown signal and then to estimate the nonzero values restricted to the active support set. First, an iterative greedy multistep procedure is developed, based on a neighborhood cooperation strategy, using batch processing on the observed data. Next, an extension of the algorithm to the online setting, based on the diffusion LMS rationale for adaptivity, is derived. Theoretical analysis of the algorithms is provided, where it is shown that the batch algorithm converges to the unknown vector if a Restricted Isometry Property (RIP) holds. Moreover, the online version converges in the mean to the solution vector under some general assumptions. Finally, the proposed schemes are tested against recently developed sparsity-promoting algorithms and their enhanced performance is verified via simulation examples. © 2015 IEEE.

The task of robust linear estimation in the presence of outliers is of particular importance in signal processing, statistics and machine learning. Although the problem has been stated a few decades ago and solved using classical (considered nowadays) methods, recently, it has attracted more attention in the context of sparse modeling, where several notable contributions have been made. In the present manuscript, a new approach is considered in the framework of greedy algorithms. The noise is split into two components: a) the inlier bounded noise and b) the outliers, which are explicitly modeled by employing sparsity arguments. Based on this scheme, a novel efficient algorithm (Greedy Algorithm for Robust Denoising-GARD), is derived. GARD alternates between a least square optimization criterion and an Orthogonal Matching Pursuit (OMP) selection step that identifies the outliers. The case where only outliers are present has been studied separately, where bounds on the Restricted Isometry Property guarantee that the recovery of the signal via GARD is exact. Moreover, theoretical results concerning convergence as well as the the recovery of the support of the sparse outlier vector and derivation of error bounds in the case of additional bounded noise are discussed. Finally, we provide extensive simulations, which demonstrate the comparative advantages of the new technique. © 1991-2012 IEEE.

In this paper, an Adaptive Projected Subgradient Method (APSM) based algorithm for robust subspace tracking is introduced. A properly chosen cost function is constructed at each time instance and the goal is to seek for points, which belong to the zero level set of this function; i.e., the set of points which score a zero loss. At each iteration, an outlier detection mechanism is employed, in order to conclude whether the current data vector contains outlier noise or not. In the sequel, a sparsity-promoting greedy algorithm is employed for the outlier vector estimation allowing the purification of the corrupted data from the outlier noise, prior to any further processing. Furthermore, the case where the observation vectors are partially observed is attacked via a prediction procedure, which estimates the values of the unobserved (missing) coefficients. A theoretical analysis is carried out and the simulation experiments, within the contexts of robust subspace estimation and robust matrix completion, demonstrate the enhanced performance of the proposed scheme compared to recently developed state of the art algorithms. © 1991-2012 IEEE.

The paper presents a new framework for complex support vector regression (SVR) as well as Support Vector Machines (SVM) for quaternary classification. The method exploits the notion of widely linear estimation to model the input-out relation for complex-valued data and considers two cases: 1) the complex data are split into their real and imaginary parts and a typical real kernel is employed to map the complex data to a complexified feature space and 2) a pure complex kernel is used to directly map the data to the induced complex feature space. The recently developed Wirtinger's calculus on complex reproducing kernel Hilbert spaces is employed to compute the Lagrangian and derive the dual optimization problem. As one of our major results, we prove that any complex SVM/SVR task is equivalent with solving two real SVM/SVR tasks exploiting a specific real kernel, which is generated by the chosen complex kernel. In particular, the case of pure complex kernels leads to the generation of new kernels, which have not been considered before. In the classification case, the proposed framework inherently splits the complex space into four parts. This leads naturally to solving the four class-task (quaternary classification), instead of the typical two classes of the real SVM. In turn, this rationale can be used in a multiclass problem as a split-class scenario based on four classes, as opposed to the one-versus-all method; this can lead to significant computational savings. Experiments demonstrate the effectiveness of the proposed framework for regression and classification tasks that involve complex data. © 2012 IEEE.

Publication in the conference proceedings of EUSIPCO, Lisbon, Portugal, 2014

Publication in the conference proceedings of EUSIPCO, Lisbon, Portugal, 2014

Publication in the conference proceedings of EUSIPCO, Lisbon, Portugal, 2014

Preamble-based channel estimation in filter bank-based multicarrier (FBMC) systems using offset quadrature amplitude modulation (OQAM) has been extensively studied in the last few years, due to the many advantages this modulation scheme can offer over cyclic prefix (CP)-based orthogonal frequency division multiplexing (OFDM) and in view of the interesting challenges posed on the channel estimator by the interference effect inherent in such an FBMC system. In particular, preambles of short duration and of both the block (full) and comb (sparse) types were designed so as to minimize the channel estimation mean squared error (MSE) subject to a given transmit energy. In the light of the important role that relay-based cooperative networks are expected to play in future wireless communication systems, it is of interest to consider FBMC/OQAM, and in particular questions associated to preamble-based channel estimation, in such a context as well. The goal of this paper is to address these problems and come up with optimal solutions that extend existing results in a single relay-based cooperative network. Both low and medium frequency selective channels are considered. In addition to optimal preamble and estimator design, the equalization/detection task is studied, shedding light to a relay-generated interference effect and proposing a simple way to come over it. The reported simulation results corroborate the analysis and reveal interesting behavior with respect to channel frequency selectivity and signal-to-noise ratio. © 2014 Mavrokefalidis et al.; licensee Springer.

I-ImaS (Intelligent Imaging Sensors) is a European project which has designed and developed a new adaptive X-ray imaging system using on-line exposure control, to create locally optimized images. The I-ImaS system allows for real-time image analysis during acquisition, thus enabling real-time exposure adjustment. This adaptive imaging system has the potential of creating images with optimal information within a given dose constraint and to acquire optimally exposed images of objects with variable density during one scan. In this paper we present the control system and results from initial tests on mammographic and encephalographic images. Furthermore, algorithms for visualization of the resulting images, consisting of unevenly exposed image regions, are developed and tested. The preliminary results show that the same image quality can be achieved at 30-70% lower dose using the I-ImaS system compared to conventional mammography systems.

Publication in the conference proceedings of EUSIPCO, Marrakech, Morocco, 2013