Source Space Research Articles

Topic-Based Unsupervised and Supervised Dictionary Induction

Word translation is a natural language processing task that provides translation between the words of a source and a target language. As a task, it reduces to the induction of a bilingual dictionary, which is typically performed by aligning word embeddings of the source language to word embeddings of the target language. To date, all the existing approaches have focused on performing a single, global alignment in word embedding space. However, semantic differences between the various languages, in addition to differences in the content of the corpora used for training the word embeddings, can hinder the effectiveness of such a global alignment. For this reason, in this article we propose conducting the alignment between the source and target embedding spaces by multiple mappings at topic level. The experimental results show that our approach has been able to achieve an average accuracy improvement of +3.30 percentage points over a state-of-the-art approach in unsupervised dictionary induction from languages as diverse as German, French, Italian, Spanish, Finnish, Turkish, and Chinese to English, and +3.95 points average improvement in supervised dictionary induction.

Unsupervised Parallel Sentences of Machine Translation for Asian Language Pairs

Parallel sentence pairs play a very important role in many natural language processing tasks, especially cross-lingual tasks such as machine translation. So far, many Asian language pairs lack bilingual parallel sentences. As collecting bilingual parallel data is very time-consuming and difficult, it is very important for many low-resource Asian language pairs. While existing methods have shown encouraging results, they rely on bilingual data seriously or have some drawbacks in an unsupervised situation. To address these issues, we propose a new unsupervised similarity calculation and dynamic selection metric to obtain parallel sentence pairs in an unsupervised situation. First, our method maps bilingual word embedding by postdoc adversarial training, which rotates the source space to match the target without parallel data. Then, we introduce a new cross-domain similarity adaption to obtain parallel sentence pairs. Experimental results on real-world datasets show that our model can obtain better accuracy and recall on mining parallel sentence pairs. We also show that the extracted bilingual sentence corpora can significantly improve the performance of neural machine translation.

A domain specific knowledge extraction transformer method for multisource satellite-borne SAR images ship detection

Multisource satellite-borne synthetic aperture radar (SAR) images have different probability distributions. Traditional supervised learning, consequently, cannot achieve good test performance on one novel satellite-borne SAR dataset while training a good model on another existing satellite-borne SAR dataset. In this article, a domain adaptation (DA) Transformer object detection method is proposed to solve the unlabeled multisource satellite-borne SAR image object detection problem. Unlike existing DA methods based on convolutional neural network (CNN) that focus more on multi-level local feature extraction, we choose to use Vision Transformer (ViT) Faster region CNN (FRCNN) as the baseline network to cope with the extraction of global features of SAR images. Then, two classification tokens are used to learn the mapping of different domains and fully extract domain-specific knowledge, generating two different feature spaces that rely on the original label and pseudo-label to train the source and target domains feature spaces, respectively. Besides, the pseudo-label of target domain is also refined and reconstructed by feature clustering in order to improve the accuracy of target domain knowledge. Finally, the original detection head of FRCNN is employed to detect the target domain SAR image objects. Extensive experiments on image datasets from multisource satellite-borne SAR such as Gaofen-3, TerraSAR-X, Sentinel-1, and RadarSAT-2 show that compared to the other state of the art (SOTA) methods, the proposed method can achieve the greatest object detection accuracy. Especially, taking the recently proposed Transformer-based method as an example, our method has more than 5% improvement in accuracy and more than 16% reduction in training time.

Open Set Domain Adaptation With Soft Unknown-Class Rejection.

The goal of domain adaptation (DA) is to train a good model for a target domain, with a large amount of labeled data in a source domain but only limited labeled data in the target domain. Conventional closed set domain adaptation (CSDA) assumes source and target label spaces are the same. However, this is not quite practical in real-world applications. In this work, we study the problem of open set domain adaptation (OSDA), which only requires the target label space to partially overlap with the source label space. Consequently, the solution to OSDA requires unknown classes detection and separation, which is normally achieved by introducing a threshold for the prediction of target unknown classes; however, the performance can be quite sensitive to that threshold. In this article, we tackle the above issues by proposing a novel OSDA method to perform soft rejection of unknown target classes and simultaneously match the source and target domains. Extensive experiments on three standard datasets validate the effectiveness of the proposed method over the state-of-the-art competitors.

Fair Numerical Algorithm of Coset Cardinality Spectrum for Distributed Arithmetic Coding.

As a typical symbol-wise solution of asymmetric Slepian-Wolf coding problem, Distributed Arithmetic Coding (DAC) non-linearly partitions source space into disjoint cosets with unequal sizes. The distribution of DAC coset cardinalities, named the Coset Cardinality Spectrum (CCS), plays an important role in both theoretical understanding and decoder design for DAC. In general, CCS cannot be calculated directly. Instead, a numerical algorithm is usually used to obtain an approximation. This paper first finds that the contemporary numerical algorithm of CCS is theoretically imperfect and does not finally converge to the real CCS. Further, to solve this problem, we refine the original numerical algorithm based on rigorous theoretical analyses. Experimental results verify that the refined numerical algorithm amends the drawbacks of the original version.

A Robust eLORETA Technique for Localization of Brain Sources in the Presence of Forward Model Uncertainties.

In this paper, we present a robust version of the well-known exact low-resolution electromagnetic tomography (eLORETA) technique, named ReLORETA, to localize brain sources in the presence of different forward model uncertainties. We first assume that the true lead field matrix is a transformation of the existing lead field matrix distorted by uncertainties and propose an iterative approach to estimate this transformation accurately. Major sources of the forward model uncertainties, including differences in geometry, conductivity, and source space resolution between the real and simulated head models, and misaligned electrode positions, are then simulated to test the proposed method. ReLORETA and eLORETA are applied to simulated focal sources in different regions of the brain and the presence of various noise levels as well as real data from a patient with focal epilepsy. The results show that ReLORETA is considerably more robust and accurate than eLORETA in all cases. Having successfully dealt with the forward model uncertainties, ReLORETA proved to be a promising method for real-world clinical applications. eLORETA is one of the localization techniques that could be used to study brain activity for medical applications such as determining the epileptogenic zone in patients with medically refractory epilepsy. However, the major limitation of eLORETA is sensitivity to the uncertainties in the forward model. Since this problem can substantially undermine its performance in real-world applications where the exact lead field matrix is unknown, developing a more robust method capable of dealing with these uncertainties is of significant interest.

Analysis of Health Care System Development in the Regions amidst the Economic Inclusiveness and Social Determinants of Health

The article proposes a neural network-based Kohonen's self-organized maps cluster analysis of Ukraine’s health care system at regional level. At analysis, economic patterns and social determinants of health are considered. The research aims to estimate regional security at the public health level. For that, behavioral and social patterns determine a regions’ potential resistance to public health risks. The authors identify the strengths and weaknesses of each region and assess the effectiveness of health care as it is provided. Interestingly, the clustering algorithm fits multidimensional space design into spaces with a lower dimension. Additionally, similar vectors in the source space appear closely on the resulting map. The algorithm design, stages of evaluation, and input groups of indicators by components are described. The data set reflects the 22 regions of Ukraine. The rationing of indicators is calculated to make the data comparable. Data are checked for quality, sparsity, duplicates, and inconsistencies. Five clusters are generated based on development of patterns within regions as well as the information value of healthcare-related socio-economic indicators. The residents of regions that belong to the first cluster systematically assess their health. Demographically, these residents are more physically active compared with residents in clusters of other regions. Findings also indicate that residents in the first cluster monitor their nutrition. The second cluster is informative on residents’ behavioral components. In the third cluster are grouped regions with financially secure residents. The fourth cluster includes leader regions. The fifth cluster includes outsider regions. The proposed model can easily fit to new data, to identify new patterns and to graphically represent new results. The model can also analyze computationally complex approach based on a complete set of multidirectional indicators relating to the country's medical system at a state of risk. Moreover, this cluster-based approach can identify areas that require increased attention by state public health agencies.

Source space connectomics of neurodegeneration: One-metric approach does not fit all

Brain functional connectivity in dementia has been assessed with dissimilar EEG connectivity metrics and estimation procedures, thereby increasing results' heterogeneity. In this scenario, joint analyses integrating information from different metrics may allow for a more comprehensive characterization of brain functional interactions in different dementia subtypes. To test this hypothesis, resting-state electroencephalogram (rsEEG) was recorded in individuals with Alzheimer's Disease (AD), behavioral variant frontotemporal dementia (bvFTD), and healthy controls (HCs). Whole-brain functional connectivity was estimated in the EEG source space using 101 different types of functional connectivity, capturing linear and nonlinear interactions in both time and frequency-domains. Multivariate machine learning and progressive feature elimination was run to discriminate AD from HCs, and bvFTD from HCs, based on joint analyses of i) EEG frequency bands, ii) complementary frequency-domain metrics (e.g., instantaneous, lagged, and total connectivity), and iii) time-domain metrics with different linearity assumption (e.g., Pearson correlation coefficient and mutual information). <10% of all possible connections were responsible for the differences between patients and controls, and atypical connectivity was never captured by >1/4 of all possible connectivity measures. Joint analyses revealed patterns of hypoconnectivity (patients<HCs) involving convergent temporo-parieto-occipital regions in AD, and fronto-temporo-parietal areas in bvFTD. Hyperconnectivity (patients>HCs) in both groups was mainly identified in frontotemporal regions. These atypicalities were differently captured by frequency- and time-domain connectivity metrics, in a bandwidth-specific fashion. The multi-metric representation of source space whole-brain functional connectivity evidenced the inadequacy of single-metric approaches, and resulted in a valid alternative for the selection problem in EEG connectivity. These joint analyses reveal patterns of brain functional interdependence that are overlooked with single metrics approaches, contributing to a more reliable and interpretable description of atypical functional connectivity in neurodegeneration.

Source-Space Brain Functional Connectivity Features in Electroencephalogram-Based Driver Fatigue Classification

This study examined the brain source space functional connectivity from the electroencephalogram (EEG) activity of 48 participants during a driving simulation experiment where they drove until fatigue developed. Source-space functional connectivity (FC) analysis is a state-of-the-art method for understanding connections between brain regions that may indicate psychological differences. Multi-band FC in the brain source space was constructed using the phased lag index (PLI) method and used as features to train an SVM classification model to classify driver fatigue and alert conditions. With a subset of critical connections in the beta band, a classification accuracy of 93% was achieved. Additionally, the source-space FC feature extractor demonstrated superiority over other methods, such as PSD and sensor-space FC, in classifying fatigue. The results suggested that source-space FC is a discriminative biomarker for detecting driving fatigue.

Proprioceptive response strength in the primary sensorimotor cortex is invariant to the range of finger movement

Proprioception is the sense of body position and movement that relies on afference from the proprioceptors in muscles and joints. Proprioceptive responses in the primary sensorimotor (SM1) cortex can be elicited by stimulating the proprioceptors using evoked (passive) limb movements. In magnetoencephalography (MEG), proprioceptive processing can be quantified by recording the movement evoked fields (MEFs) and movement-induced beta power modulations or by computing corticokinematic coherence (CKC) between the limb kinematics and cortical activity. We examined whether cortical proprioceptive processing quantified with MEF peak strength, relative beta suppression and rebound power and CKC strength is affected by the movement range of the finger.MEG activity was measured from 16 right-handed healthy volunteers while movements were applied to their right-index finger metacarpophalangeal joint with an actuator. Movements were either intermittent, every 3000 ± 250 ms, to estimate MEF or continuous, at 3 Hz, to estimate CKC. In both cases, 4 different ranges of motion of the stimuli were investigated: 15, 18, 22 and 26 mm for MEF and 6, 7, 9 and 13 mm for CKC. MEF amplitude, relative beta suppression and rebound as well as peak CKC strength at the movement frequency were compared between the movement ranges in the source space. Inter-individual variation was also compared between the MEF and CKC strengths.As expected, MEF and CKC responses peaked at the contralateral SM1 cortex. MEF peak, beta suppression and rebound and CKC strengths were similar across all movement ranges. Furthermore, CKC strength showed a lower degree of inter-individual variation compared with MEF strength.Our result of absent modulation by movement range in cortical responses to passive movements of the finger indicates that variability in movement range should not hinder comparability between different studies or participants. Furthermore, our data indicates that CKC is less prone to inter-individual variability than MEFs, and thus more advantageous in what pertains to statistical power.

Geminga SNR: Possible Candidate of Local Cosmic-Ray Factory (II)

Accurate measurements of the energy spectrum and anisotropy can help us discover local cosmic-ray accelerators. Our recent works have shown that spectral hardening above 200 GeV in the energy spectra and transition of large-scale anisotropy at ∼100 TeV are of an unifying origin. Less than 100 TeV, both spectral hardening and anisotropy explicitly indicate the dominant contribution from nearby sources. Recent observations of CR anisotropy suggest that this phase is consistent with the locally regular magnetic field (LRMF) of the interstellar boundary explorer (IBEX) below 100 TeV. In this work, we further investigate the parameter space of sources allowed by the observational energy spectra and amplitude and phase of dipole anisotropy. To obtain the best-fit source parameters, a numerical algorithm is to compute the parameter posterior distributions based on Bayesian inference. We found that by combining the observations of the energy spectrum and anisotropy, the parameters of the model can be well constrained. The LRMF and the effect of the corresponding anisotropic diffusion are considered in this work. Finally, the phase results’ right ascension (R.A.)=3.2 h below 100 TeV was obtained by fitting, which is in general agreement with the experimental observations. Since the Geminga SNR is very close to the mean of the fitted parameters, it could be a candidate for a local cosmic-ray accelerator.

Opposite Modulation of the NMDA Receptor by Glycine and S-Ketamine and the Effects on Resting State EEG Gamma Activity: New Insights into the Glutamate Hypothesis of Schizophrenia.

NMDA-receptor hypofunction is increasingly considered to be an important pathomechanism in schizophrenia. However, to date, it has not been possible to identify patients with relevant NMDA-receptor hypofunction who would respond to glutamatergic treatments. Preclinical models, such as the ketamine model, could help identify biomarkers related to NMDA-receptor function that respond to glutamatergic modulation, for example, via activation of the glycine-binding site. We, therefore, aimed to investigate the effects of opposing modulation of the NMDA receptor on gamma activity (30-100 Hz) at rest, the genesis of which appears to be highly dependent on NMDA receptors. The effects of subanesthetic doses of S-ketamine and pretreatment with glycine on gamma activity at rest were examined in twenty-five healthy male participants using 64-channel electroencephalography. Psychometric scores were assessed using the PANSS and the 5D-ASC. While S-ketamine significantly increased psychometric scores and gamma activity at the scalp and in the source space, pretreatment with glycine did not significantly attenuate any of these effects when controlled for multiple comparisons. Our results question whether increased gamma activity at rest constitutes a suitable biomarker for the target engagement of glutamatergic drugs in the preclinical ketamine model. They might further point to a differential role of NMDA receptors in gamma activity generation.

Earthquake Phase Association with Graph Neural Networks

ABSTRACTSeismic phase association connects earthquake arrival-time measurements to their causative sources. Effective association must determine the number of discrete events, their location, and origin times, and it must differentiate real arrivals from measurement artifacts. The advent of deep-learning (DL) pickers, which provide high rates of picks from closely overlapping small-magnitude earthquakes, motivates revisiting the phase association problem and approaching it using the methods of DL. We have developed a graph neural network associator that simultaneously predicts both source space–time localization, and discrete source-arrival association likelihoods. The method is applicable to arbitrary geometry, time-varying seismic networks of hundreds of stations, and is robust to high rates of sources and input picks with variable noise and quality. Our Graph Earthquake Neural Interpretation Engine (GENIE) uses one graph to represent the station set and another to represent the spatial source region. GENIE learns relationships from data in this combined representation that enable it to determine robust source and source-arrival associations. We train on synthetic data, and test our method on real data from the northern California seismic network using input generated by the PhaseNet DL phase picker. We successfully re-detect ∼96% of all events M &amp;gt;1 reported by the U.S. Geological Survey (USGS) during 500 random days between 2000 and 2022. Over a 100-day continuous interval of processing in 2017–2018, we detect ∼4× the number of events reported by the USGS. Our new events have small-magnitude estimates below the magnitude of completeness of the USGS catalog, and are located close to the active faults and quarries in the region. Our results demonstrate that GENIE can effectively solve the association problem under complex seismic monitoring conditions.

The Effect of Planting Arrangement of Nitrogen and Phosphorus on Yield and Physiological Traits of Guar (Cyamopsis tetragonoloba L.) in Iran

To study the effect of planting arrangement of nitrogen and phosphorus on yield and physiological traits of guar, a factorial experiment in a randomized complete block design with three replications in two cropping years 2019-20 and 2020-21 in the research center and agricultural and natural resources education of southern Kerman was implemented. Nitrogen at three levels (20, 60, and 100 kg pure nitrogen per hectare) from urea and phosphorus fertilizer source at two levels (zero and 75 kg P2O5 source per hectare) from monopotassium phosphate fertilizer source and planting row spacing at three levels (30, 45 and 60 cm) equivalent to 47, 31 and 24 plants per square meter, plant density were considered. The results showed that nitrogen, phosphorus, and planting density significantly affected all traits. Planting row spacing affected all characteristics except seed gum and phosphorus levels were effective on all traits except glycine betaine and leaf sodium. The highest yield components and seed gum were obtained from N3P2D3 treatment (application of 60kg/ha nitrogen, 75 kg/ha phosphorus, and 60 cm row spacing) which was no different from N2P1D3 treatment and the highest leaf proline content was related to N2P2D3 treatment and the highest glycine content was betaine and leaves. Sodium was related to N2P1D3 treatment. The highest grain yield and yield components in the first and second year of N2P2D2 treatment (application of 60 and 75 kg of nitrogen and phosphorus, respectively, with a row spacing of 45 cm) were obtained at 4300 and 4200 kg/ha, respectively, with N3P1D2 and N3P2D2 was not significantly different. The highest levels of carotenoids and chlorophyll a and b were related to N3P2D3 treatment. Based on the results of this study, it seems that this plant does not need high levels of nitrogen and phosphorus in terms of legume, but is affected by nitrogen and phosphorus for high yield and product quality. Regarding row spacing, if the goal is to achieve a high yield, row spacing of 45 cm and high quality 60 cm spacing is recommended and can be considered as a low-expectation plant in the cultivation pattern program in the climatic conditions of southern Kerman

A Cross-Space CNN With Customized Characteristics for Motor Imagery EEG Classification

The classification of motor imagery-electroencephalogram(MI-EEG)based brain-computer interface(BCI)can be used to decode neurological activities, which has been widely applied in the control of external devices. However, two factors still hinder the improvement of classification accuracy and robustness, especially in multi-class tasks. First, existing algorithms are based on a single space (measuring or source space). They suffer from the holistic low spatial resolution of the measuring space or the locally high spatial resolution information accessed from the source space, failing to provide holistic and high-resolution representations. Second, the subject specificity is not sufficiently characterized, resulting in the loss of personalized intrinsic information. Therefore, we propose a cross-space convolutional neural network (CS-CNN) with customized characteristics for four-class MI-EEG classification. This algorithm uses the modified customized band common spatial patterns (CBCSP) and duplex mean-shift clustering (DMSClustering) to express the specific rhythms and source distribution information in cross-space. At the same time, multi-view features from the time, frequency and space domains are extracted, connecting with CNN to fuse the characteristics from two spaces and classify them. MI-EEG was collected from 20 subjects. Lastly, the classification accuracy of the proposed is 96.05% with real MRI information and 94.79% without MRI in the private dataset. And the results in the BCI competition IV-2a show that CS-CNN outperforms the state-of-the-art algorithms, achieving an accuracy improvement of 1.98%, and a standard deviation reduction of 5.15%.

Tinnitus-frequency specific activity and connectivity: A MEG study

Tinnitus pathophysiology has been associated with an atypical cortical network that involves functional changes in auditory and non-auditory areas. Numerous resting-state studies have replicated a tinnitus brain network to be significantly different from healthy-controls. Yet it is still unknown whether the cortical reorganization is attributed to the tinnitus frequency specifically or if it is frequency-irrelevant. Employing magnetoencephalography (MEG), the current study aimed to identify frequency-specific activity patterns by using an individual tinnitus tone (TT) and a 500 Hz-control tone (CT) as auditory stimuli, across 54 tinnitus patients. MEG data were analyzed in a data-driven approach employing a whole-head model in source space and in sources’ functional connectivity. Compared to the CT, the event related source space analysis revealed a statistically significant response to TT involving fronto-parietal regions. The CT mainly involved typical auditory activation-related regions. A comparison of the cortical responses to a healthy control group that underwent the same paradigm rejected the alternative interpretation that the frequency-specific activation differences were due to the higher frequency of the TT. Overall, the results suggest frequency-specificity of tinnitus-related cortical patterns. In line with previous studies, we demonstrated a tinnitus-frequency specific network comprising left fronto-temporal, fronto-parietal and tempo-parietal junctions.

Decoding Natural Grasping Behaviors: Insights Into MRCP Source Features and Coupling Dynamics.

The effective decoding of natural grasping behaviors is crucial for the natural control of neural prosthetics. This study aims to investigate the decoding performance of movement-related cortical potential (MRCP) source features between complex grasping actions and explore the temporal and frequency differences in inter-muscular and cortical-muscular coupling strength during movement. Based on the human grasping taxonomy and their frequency, five natural grasping motions-medium wrap, adducted thumb, adduction grip, tip pinch, and writing tripod-were chosen. We collected 64-channel electroencephalogram (EEG) and 5-channel surface electromyogram (sEMG) data from 17 healthy participants, and projected six EEG frequency bands into source space for further analysis. Results from multi-classification and binary classification demonstrated that MRCP source features could not only distinguish between power grasp and precision grasp, but also detect subtle action differences such as thumb adduction and abduction during the execution phase. Besides, we found that during natural reach-and-grasp movement, the coupling strength from cortical to muscle is lower than that from muscle to cortical, except in the hold phase of γ frequency band. Furthermore, a 12-Hz peak of inter-muscular coupling strength was found in movement execution, which might be related to movement planning and execution. We believe that this research will enhance our comprehension of the control and feedback mechanisms of human hand grasping and contributes to a natural and intuitive control for brain-computer interface.

Experimental study on the flame merging and ceiling impingement behavior of transversely located double fire sources in an urban utility tunnel

The investigation of the fire characteristics and propagation behaviour in the urban utility tunnel is of great practical importance, especially for the conditions with double fire sources. In this work, a 1/8th scaled urban utility tunnel model was built to conduct the experiments to characterize the flame merging and ceiling impingement behaviour of transversely located double fire sources. Two rectangular fire sources with the same dimension were used, and their heat release rate (HRR) and fire source spacings were varied to consider the typical scenarios. Results show that for the smaller HRR, the tunnel ceiling and sidewalls have little influence on the flame merging of double fire sources. With the increase of HRR, the flame gradually impinges on the tunnel ceiling and forms a stable ceiling jet flame extension, and the ceiling and sidewalls of the tunnel can promote the merging of the flames of the double fire sources. Besides, with the increase of the fire source spacings, the flame merging probability P m can be divided into three stages, that is, (i) complete merging stage, (ii) intermittent merging stage and (iii) complete separation stage. On this basis, the predicting relation of P m was obtained by using the piecewise function.

Helical graphs to visualize the NDVI temporal variation of forest vegetation in an open source space

Global change caused by human activity (e.g., land fragmentation, deforestation, pollution, anthropization of natural landscapes) has several effects on the biomes of the Earth, leading to alterations in the functioning of ecological systems. In this context, remote sensing represents an important tool to assess ecosystem changes, as it allows to collect a huge amount of data at different temporal and spatial resolutions concerning various compartments of the Earth system (land, ocean, atmosphere, and cryosphere). This information can be used to estimate precipitation patterns, global temperatures, snow cover and aerosol concentrations. The aim of this work is to exploit this wide availability of data to display the ecosystem changes using a new visualization method: the helical graphs. The helical graphs represent the change of a variable over time, reporting on the y-axis its moving averages and on the x-axis its rates of change.These new charts were tested on the NDVI index retrieved from Google Earth Engine (https://earthengine.google.com/) to visualize trends on selected biomes of the Earth (tropical and boreal forests). The results show that the helical graphs are a useful tool to highlight trends that might not be easily detected in a time series. In conclusion, the helical graphs can have a lot of application in ecology, especially exploiting the wide amount of data available thanks to the remote sensing.

Novel advances in the characterization of dementia subtypes: from multi‐feature classification to whole brain computational models

AbstractBackgroundBrain functional connectivity analyses derived from electroencephalography (EEG) provides relevant information for classification of dementia subtypes. The predictive strength of classification tools can be benefit from integrative, multi‐feature analysis of EEG which result in composite metric of functional connectivity. Additionally, significant improvement can be obtained when connectivity analyses consider the dependence between groups of regions as a whole system (high order interactions), instead of conceiving the network as a collection of pairwise interactions.MethodFive minutes, resting‐state EEG (rsEEG) was recorded from healthy controls (HC, n=42), and participants diagnosed with either behavioral variant frontotemporal dementia (bvFTD, n=19) or Alzheimer Disease (AD, n=33). EEG source localization analyses were conducted. Whole brain functional connectivity (82 anatomic compartments, AAL atlas) was analyzed using 17 metrics to capture different statistical dependence between source space rsEEG time series, in both time‐, and frequency‐domains. Furthermore, high‐order, non‐lineal statistical interdependencies between group of brain regions were assessed using multivariate information theory. Gradient boosting classifiers with Bayesian optimization, harmonization, and feature elimination were implemented.ResultClassification of dementia subtypes, using a particular frequency‐domain connectivity metric, displayed better performances as features for classification comprised information from multiple EEG frequency bands. Likewise, classification systems including information from several connectivity metrics outperformed those achieved with individual metrics. Dementia classification conducted with connectivity metrics derived from mutual information significantly improved as order of interaction (number of brain regions consider as a network) increased. This latter approach boosted multimodal (fMRI‐EEG) classification of dementias based on brain functional connectivity, and served as input for whole‐brain computational models describing the pathophysiology of neurodegenerative diseases.ConclusionIntegrative, composite metrics of connectivity, and connectivity analyses of source localized rsEEG based in high order interactions capture relevant information to discriminate dementia subtypes, and provide a reliable and interpretable description of brain functional connectivity and neurodegeneration.