Errors In Data Research Articles

Data Validation Utilizing Expert Knowledge and Shape Constraints

Data validation is a primary concern in any data-driven application, as undetected data errors may negatively affect machine learning models and lead to suboptimal decisions. Data quality issues are usually detected manually by experts, which becomes infeasible and uneconomical for large volumes of data. To enable automated data validation, we propose “shape constraint-based data validation”, a novel approach based on machine learning that incorporates expert knowledge in the form of shape constraints. Shape constraints can be used to describe expected (multivariate and nonlinear) patterns in valid data, and enable the detection of invalid data which deviates from these expected patterns. Our approach can be divided into two steps: (1) shape-constrained prediction models are trained on data, and (2) their training error is analyzed to identify invalid data. The training error can be used as an indicator for invalid data because shape-constrained models can fit valid data better than invalid data. We evaluate the approach on a benchmark suite consisting of synthetic datasets, which we have published for benchmarking similar data validation approaches. Additionally, we demonstrate the capabilities of the proposed approach with a real-world dataset consisting of measurements from a friction test bench in an industrial setting. Our approach detects subtle data errors that are difficult to identify even for domain experts.

Wave-CapNet: A Wavelet Neuron-based Wi-Fi Sensing Model for Human Identification

Gait is regarded as a unique feature for identifying people, and gait recognition is the basis of various customized services of the IoT. Unlike traditional techniques for identifying people, the Wi-Fi-based technique is unconstrained by illumination conditions and such that it eliminates the need for dense, specialized sensors and wearable devices. Although deep learning-based sensing models are conducive to the development of Wi-Fi-based identification, the latter technique relies on a large amount of data and requires a long training time, where this limits the scope of its use for identifying people. In this study, we propose a Wi-Fi sensing model called Wave-CapNet for human identification. We use data processing to eliminate errors in the raw data so that the model can extract the characteristics in channel state information (CSI). We also design a dedicated adaptive wavelet neural network to extract representative features from Wi-Fi signals with only a few epochs of training and a small number of parameters. Experiments show that it can identify human gait with an average accuracy of 99%. Moreover, it can achieve an average accuracy of 95% by using only 10% of the data and fewer than five epochs and outperforms state-of-the-art (SOTA) methods.

The application of artificial intelligence in Unmanned Underwater Vehicle communication systems

Artificial Intelligence (AI) has the potential to significantly enhance decision-making techniques in Unmanned Underwater Vehicle (UUV) communication systems by utilizing real-time sensor feedback and environmental data. To address the challenges posed by high data volume and error rates, we propose an advanced methodology known as the Variational Deep Network (VDN). This method integrates Variational Modal Decomposition (VMD) algorithms with the Deep Convolutional Neural Network (DCNN) to create a realizable AI decision-making system. In conjunction with the DCNN and VMD algorithms, our research focuses on accurately identifying received signals under data constraints. Considering the spectral efficiency and robustness, the Filter-Bank Multicarrier (FBMC) technology have been employed in this paper. Our results demonstrate that the VDN significantly reduces the volume of data needed for neural network training and effectively recognizes FBMC signals. These insights confirm the potential of VDN decision-making systems to drive advancements in UUV communication technologies.

Active Learning for Data Quality Control: A Survey

Data quality plays a vital role in scientific research and decision-making across industries. Thus it is crucial to incorporate the data quality control (DQC) process, which comprises various actions and operations to detect and correct data errors. The increasing adoption of machine learning (ML) techniques in different domains has raised concerns about data quality in the ML field. On the other hand, ML’s capability to uncover complex patterns makes it suitable for addressing challenges involved in the DQC process. However, supervised learning methods demand abundant labeled data, while unsupervised learning methods heavily rely on the underlying distribution of the data. Active learning (AL) provides a promising solution by proactively selecting data points for inspection, thus reducing the burden of data labeling for domain experts. Therefore, this survey focuses on applying AL to DQC. Starting with a review of common data quality issues and solutions in the ML field, we aim to enhance the understanding of current quality assessment methods. We then present two scenarios to illustrate the adoption of AL into the DQC systems on the anomaly detection task, including pool-based and stream-based approaches. Finally, we provide the remaining challenges and research opportunities in this field.

Influence of Supraglacial Lakes on Accuracy of Inversion of Greenland Ice Sheet Surface Melt Data in Different Passive Microwave Bands

The occurrence of Supraglacial Lakes (SGLs) may influence the signals acquired with microwave radiometers, which may result in a degree of uncertainty when employing microwave radiometer data for the detection of surface melt. Accurate monitoring of surface melting requires a reasonable assessment of this uncertainty. However, there is a scarcity of research in this field. Therefore, in this study, we computed surface melt in the vicinity of Automatic Weather Stations (AWSs) by employing Defense Meteorological Satellite Program (DMSP) Ka-band data and Soil Moisture and Ocean Salinity (SMOS) satellite L-band data and extracted SGL pixels by utilizing Sentinel-2 data. A comparison between surface melt results derived from AWS air temperature estimates and those obtained with remote sensing inversion in the two different bands was conducted for sites below the mean snowline elevation during the summers of 2016 to 2020. Compared with sites with no SGLs, the commission error (CO) of DMSP morning and evening data at sites where these water bodies were present increased by 36% and 30%, respectively, and the number of days with CO increased by 12 and 3 days, respectively. The omission error (OM) of SMOS morning and evening data increased by 33% and 32%, respectively, and the number of days with OM increased by 17 and 21 days, respectively. Identifying the source of error is a prerequisite for the improvement of surface melt algorithms, for which this study provides a basis.

Characterization of a novel pulse normalization technology for beam scanning of small fields without a reference chamber.

A novel pulse normalization technology enabling the acquisition of low noise beam data without the use of a physical reference chamber has recently been commercially released. The purpose of this study was to characterize the use of this technology for beam scanning of small fields required in the commissioning of a stereotactic radiotherapy program. Three detectors (Edge diode, microDiamond, PinPoint) were used to acquire beam data under three conditions: with a reference chamber, with pulse normalization and no reference chamber (PN), and without pulse normalization and no reference chamber (nPN). Percent depth dose (PDD) scans were acquired for 0.5, 1.0, 2.0, and 3.0 cm2 field sizes and profiles were acquired at 1.4, 10, and 30cm depths using continuous scanning. The coefficient of variation (CoV) was calculated for all beam data to compare signal-to-noise and gamma comparisons (1%, 1mm) were calculated of the PN and nPN scans compared to the reference data. Average 95th percentile CoV values were similar for all detectors across conditions, with PN data being comparable to reference data and minor increases observed for nPN data. Mean gamma pass rates for PN PDD scans exceeded 98% for all detectors. Profile gamma pass rates were 100% for all detectors at 1.4 and 10cm depth. At 30cm depth, profiles acquired with the PinPoint and microDiamond detectors had lower mean gamma pass rates than the Edge, at 95% and 95.7%, respectively. A novel pulse normalization technology was demonstrated to be effective for acquiring beam profiles and PDDs for small fields without the use of a physical reference chamber. Limitations in how the method is implemented led to some errors in data acquired using lower sensitivity detectors. When used with a diode, pulse normalization produced equivalent scans to those acquired with a reference chamber.

Concurrent validity and test–retest reliability of VALD ForceDecks strength, balance, and movement assessment tests

ObjectiveTo evaluate the concurrent validity and test–retest reliability of common movement, strength, and balance tests using portable uniaxial dual force plates. DesignRepeated measures cross-sectional study. MethodsSixteen healthy individuals participated in two testing sessions, where they performed 12 different movement, strength, and balance tests. Vertical ground reaction forces (GRF) and centre of pressure (COP) data were collected using the VALD ForceDecks simultaneously with ground-embedded laboratory force plates. Concurrent validity was assessed by quantifying the agreement between force plate systems using root mean square error for raw time-series data and Bland–Altman plots for discrete metrics. Test–retest reliability was assessed using intraclass correlation coefficients (ICC) and minimal detectable changes. ResultsForceDecks recorded vertical GRF and COP with high accuracy compared to laboratory force plates. The mean bias between systems was negligible (<2 N or 0.1 mm), with small limits of agreement (<5 N or 1 mm). Of all possible discrete force and COP metrics, 530/674 (79 %) showed good or excellent validity (<10 % difference) and 611/773 (79 %) had good or excellent reliability (ICC > 0.75). ForceDecks reliability was similar to laboratory force plates (<0.07 ICC median difference for all metrics). ConclusionsPortable uniaxial force plates record highly accurate vertical GRF and COP during a range of movement, strength, and balance tests. The VALD ForcDecks are a valid and reliable alternative to laboratory force plates when strict standardized testing and data analysis procedures are followed. Users should be aware of the validity and reliability characteristics of the tests and metrics they choose.

Development of a multi-level factorial data assimilation approach for groundwater contaminant sources identification

The ensemble Kalman filter (EnKF) is recognized as a widely used approach for groundwater contamination sources (GCSs) parameters identification. The identification accuracy is affected by the quantification of sources of uncertainty (such as random disturbances on model inputs and measurements) in the assimilation process, but few studies have revealed the contributions of different sources of uncertainty on identification accuracy. To this end, this study developed an innovative multi-level factorial data assimilation (MFDA) approach, which integrates ensemble Kalman filtering and multi-level factorial analysis, to comprehensively reveal the impact of sources of uncertainty on the GCSs parameters identification accuracy for the first time. Not only did it explore the direct impact of individual sources of uncertainty on the GCSs parameters identification accuracy, more importantly, it also revealed the composite ones from multi-parameter interactions among these sources of uncertainty. The developed MFDA approach quantified the contributions of multiple sources of uncertainty, including model input data error (i.e. precipitation), model parameter error (i.e. hydraulic conductivity), contaminant concentration measurements error, and ensemble size. And the feasibility of the developed approach was verified by two hypothetical numerical case studies. The results indicated that: i) it was critically important to reduce the contamination concentration measurement errors to improve GCSs parameter identification accuracy. ii) model parameter and model input data were not sensitive to the identification accuracy. iii) the interaction between sources of uncertainty significantly affects the identification accuracy, which illustrated the interaction cannot be ignored in the identification process. Overall, this study highlights that the developed MFDA approach represents a promising solution for improving GCSs parameters identification accuracy.

Simulation application of sensors based on Kalman filter algorithm in student psychological crisis prediction model

The mental health and psychological crisis of some Chinese college students today are extremely abnormal, which has attracted the attention of many relevant personnel. Due to various external reasons, the psychological construction of Chinese college students is very pessimistic. Kalman filter is a regression calculation method for processing data. The standard calculation of this filter has the smallest data error, so that relevant data can be recursive. Within the relevant time domain, this calculation method can select suitable filters to accurately calculate high-dimensional and low-dimensional system data. This paper mainly solves some problems encountered, thus proving the effectiveness of Kalman filter calculation method. Finally, we can get the advantages and disadvantages of these filter systems, so as to improve these disadvantages, and finally improve the Rate of convergence of this calculation method. Through the corresponding experimental results, we can see that these calculation methods are correct. By analyzing these data, the analysis results show that this calculation method can effectively predict students' mental health problems, and the designed system can reduce the occurrence of psychological crisis events among college students.

Multivariate time series anomaly detection via separation, decomposition, and dual transformer-based autoencoder

Multivariate time series usually have entangled temporal patterns and various anomaly types. Meanwhile, they often contain both continuous and discrete features. Many existing methods directly model correlations in complex multivariate time series to conduct anomaly detection. Decomposing time series into different components, such as the overall trend and fluctuations, can contribute to better extracting semantic information and detecting anomalies. Existing decomposition-based anomaly detection methods still have several limitations. First, they directly decompose all features without considering that discrete features are unsuitable for decomposition because they do not have trends or fluctuations. Second, they adopt the same networks for different components with different characteristics, limiting their ability to extract semantic information. Moreover, due to the nature of Transformers, existing reconstruction-based methods using Transformers rarely form information bottlenecks, reducing the differentiation between the reconstruction errors of normal data and anomalies. This paper proposes a multivariate time series anomaly detection method with separation, decomposition, and dual Transformer-based autoencoder (SDDformer). Different from existing methods, SDDformer separates continuous and discrete features and only decomposes continuous features into trend and residual components. Considering the different characteristics of different components, SDDformer adopts Crossformer and the vanilla Transformer as the backbone of two different autoencoders to reconstruct the trend and residual components. Information bottlenecks are better formed using an extra token as the latent variable between the encoder and the decoder. SDDformer regards reconstructing a discrete feature as a classification task and calculates Cross-Entropy as its reconstruction error. Three different metrics are adopted in this paper to compare SDDformer with a variety of typical anomaly detection methods on public data sets, and the experimental results prove that SDDformer can achieve state-of-the-art performance.

Retrospective correction of second-order concomitant fields in 3D axial stack-of-spirals imaging on a high-performance gradient system.

MRI using 3D stack-of-spirals (SoS) readout on a high-performance gradient system is subject to strong second-order, spatially varying concomitant fields, which can lead to signal dropout and blurring artifacts that become more significant at locations farther from the gradient isocenter. A method for compensating for second-order concomitant fields in 3D axial SoS image reconstruction is described. We retrospectively correct for second-order concomitant field-induced phase error in the 3D SoS data by slice-dependent k-space phase compensation based on the nominal spiral readout trajectories. The effectiveness of the method was demonstrated in phantom and healthy volunteer scans in which 3D pseudo-continuous arterial spin labeling imaging was performed with SoS fast spin-echo readout at 3 T. Substantial reduction in blurring was observed with the proposed method. In phantom scans, blurring was reduced by about 53% at 98 mm from the gradient isocenter. In the in vivo 3D pseudo-continuous arterial spin labeling scans, differences of up to 10% were observed at 78 mm from the isocenter, especially around the white-matter and gray-matter interfaces, between the corrected and uncorrected proton density images, perfusion-weighted images, and cerebral blood flow maps. The described retrospective correction method provides a means to correct erroneous phase accruals due to second-order concomitant fields in 3D axial stack-of-spirals imaging.

A Network Enhancement Method to Identify Spurious Drug-Drug Interactions.

As medical safety and drug regulation gain heightened attention, the detection of spurious drug-drug interactions (DDI) has become key in healthcare. Although current research using graph neural networks (GNNs) to predict DDI has shown impressive results, it often fails to account for false DDI in the constructed DDI networks. Such inaccuracies caused by data errors, false alarms, or incorrect drug details can skew the network's structure and hinder the accuracy of GNN-based predictions. To tackle this challenge, we propose ANSM, a network-enhancement method specifically designed to identify and attenuate spurious links between drugs for ensuring the accuracy of DDI networks. ANSM integrates three key components: the feature extractor, the network optimizer, and the discriminative classifier. The feature extractor captures local structural features from drug node pairs, while the network optimizer leverages network information to improve feature extraction and reduce the impact of spurious DDI links. The discriminative classifier then identifies potential spurious links. Experimental results demonstrate that ANSM outperforms state-of-the-art methods in identifying spurious DDI.

Effects of Equatorial Plasma Bubbles on Multi-GNSS Signals: A Case Study over South China

Equatorial plasma bubbles (EPBs) occur frequently in low-latitude areas and have a non-negligible impact on navigation satellite signals. To systematically analyze the effects of a single EPB event on multi-frequency signals of GPS, Galileo, GLONASS, and BDS, all-sky airglow images over South China are jointly used to visually determine the EPB structure and depletion degree. The results reveal that scintillations, or GNSS signal fluctuations, are directly linked to EPBs and that the intensity of scintillation is positively correlated with the airglow depletion intensity. The center of the airglow depletion often corresponds to stronger GNSS scintillation, while the edge of the bubble, which is considered to have the largest density gradient, corresponds to relatively smaller scintillation instead. This work also systematically analyzes the responses of multi-constellation and multi-frequency signals to EPBs. The results show that the L2 and L5 frequencies are more susceptible than the L1 frequency is. For different constellations, Galileo’s signal has the best tracking stability during an EPB event compared with GPS, GLONASS, and BDS. The results provide a reference for dual-frequency signal selection in precise positioning or TEC calculation, that is, L1C and L2L for GPS, L1C and L5Q for Galileo, L1P and L2C for GLONASS, and L1P and L5P for BDS. Notably, BDS-2 is significantly weaker than BDS-3. And inclined geosynchronous orbit (IGSO) satellites have abnormal data error rates, which should be related to the special signal path trajectory of the IGSO satellite.

High resolution chlorophyll-a in-situ fluorescence sensors versus in-vitro chlorophyll-a measurements in mesocosms with contrasting nutrient and temperature treatments

Harmful algal blooms (HABs) are a significant threat to freshwater ecosystems, and monitoring for changes in biomass is therefore important. Fluorescence in-situ sensors enable rapid and high frequency real-time data collection and have been widely used to determine chlorophyll-a (Chla) concentrations that are used as an indicator of the total algal biomass. However, conversion of fluorescence to equivalent Chla concentrations is often complicated due to biofouling, phytoplankton composition and the type of equipment used. Here, we validated measurements from 24 Chla and 12 phycocyanin (cyanobacteria indicator) fluorescence in-situ sensors (Cyclops-7F, Turner Designs) against spectrophotometrically (in-vitro) determined Chla and tested a data-cleaning procedure for eliminating data errors and impacts of non-photochemical quenching. The test was done across a range of freshwater plankton communities in 24 mesocosms (i.e. experimental tanks) with a 2x3 (high and low nutrient x ambient, IPCC-A2 and IPCC-A2+50% temperature scenarios) factorial design. For most mesocosms (tanks), we found accurate (r2 ≥ 0.7) calibration of in-situ Chla fluorescence data using simple linear regression. An exception was tanks with high in-situ phycocyanin fluorescence, for which multiple regressions were employed, which increased the explained variance by &gt;16%. Another exception was the low Chla concentration tanks (r2 &lt; 0.3). Our results also show that the high frequency in-situ fluorescence data recorded the timing of sudden Chla variations, while less frequent in-vitro sampling sometimes missed these or, when recorded, the duration of changes was inaccurately determined. Fluorescence in-situ sensors are particularly useful to detect and quantify sudden phytoplankton biomass variations through high frequency measurements, especially when using appropriate data-cleaning methods and accounting for factors that can impact the fluorescence readings.

Artificial Intelligence Approach in Aerospace for Error Mitigation

Many of the reports created at assembly lines, where all components of an aircraft are installed, frequently indicate that errors threaten safety. The proposed methodology in this study evaluates error prediction and risk mitigation to prevent failures and their consequences. The results linked to a typical electrical harness manufacture of a military aircraft estimated reductions of 93% in time and 90% in error during the creation of engineering manufacturing processes using AI techniques. However, traditional risk assessments methods struggle to identify and mitigate errors effectively. Thus, developing an advanced methodology to ensure systems safety is needed. This paper addresses how innovative AI technology solutions can overcome these challenges, mitigate error risks, and enhance safety in aerospace. Technologies, such as artificial intelligence, predictive algorithms, machine learning, and automation, can play a key role in enhancing safety. The aim of this study is to develop a model that considers the factors that can potentially contribute to error creation, through an artificial intelligence (AI) approach. The specific AI techniques used such as support vector machine, random forest, logistic regression, K-nearest neighbor, and XGBoost (Python 3.8.5) show good performance for use in error mitigation. We have compared the modeled values obtained in this study with the experimental ones. The results confirm that the best metrics are obtained by using support vector machine and logistic regression. The smallest deviation between the measured and modeled values for these AI methods do not exceed 5%. Furthermore, using advancements in machine learning methods can enhance error mitigation in aerospace. The use of AutoML can play a key role in automatically finding an appropriate model which provides the best performance metrics and therefore the most reliable forecast for data prediction and error mitigation.

Innovative approach to predict soil moisture using the backpropagation‐Elman neural network

AbstractAccurate moisture prediction aids in taking early preventive measures to minimize damage caused by severe storms, droughts, and floods. The study collected 24 types of statistical data from the Xilingol Grassland, Inner Mongolia, spanning from 2011 to 2021. By integrating Elman, backpropagation (BP), and soil moisture prediction model (Elman‐BP) neural networks, soil moisture to 2 m was predicted. The soil moisture mechanism model analysis revealed that soil moisture is related to water supply rather than directly correlated with time. Based on these findings, a two‐stage soil moisture prediction model was established. In the first stage, the Elman neural network model predicted future precipitation data based on observed values. In the second stage, soil moisture was predicted using a BP neural network model based on predicted precipitation data. Comparing the mean squared errors of training and prediction data across the three models revealed that the Elman‐BP neural network had superior predictive accuracy. The maximum relative errors for the trained Elman neural network, BP neural network, and Elman‐BP model were 31.35%, 20.98%, and 3.94%. Overall, results showed a link between soil moisture and water supply and that soil moisture decreased with increasing temperatures. Soil moisture at 10‐ and 40‐cm depths was increased with increasing rainfall. Furthermore, a delay exists between the 40‐ and 100‐cm soil depths. Moreover, soil moisture stability increased with depth.

Global Navigation Satellite System Receiver Positioning in Harsh Environments via Clock Bias Prediction by Empirical Mode Decomposition and Back Propagation Neural Network Method.

This paper proposes a novel method to improve the clock bias short-term prediction accuracy of navigation receivers then solve the problem of low positioning accuracy when the satellite signal quality deteriorates. Considering that the clock bias of a navigation receiver is equivalent to a virtual satellite, the predicted value of clock bias is used to assist navigation receivers in positioning. Consequently, a combined prediction method for navigation receiver clock bias based on Empirical Mode Decomposition (EMD) and Back Propagation Neural Network (BPNN) analysis theory is demonstrated. In view of systematic errors and random errors in the clock bias data from navigation receivers, the EMD method is used to decompose the clock bias data; then, the BPNN prediction method is used to establish a high-precision clock bias prediction model; finally, based on the clock bias prediction value, the three-dimensional positioning of the navigation receiver is realized by expanding the observation equation. The experimental results show that the proposed model is suitable for clock bias time series prediction and providing three-dimensional positioning information meets the requirements of navigation application in the harsh environment of only three satellites.

Fault diagnosis of gearboxin wind turbine based on EMD-DCGAN

INTRODUCTION: Wind turbine gearbox fault diagnosis is of great significance for the safe and stable operation of wind turbines. The accuracy of wind turbine gearbox fault diagnosis can be effectively improved by using complete wind turbine gearbox fault data and efficient fault diagnosis algorithms.A wind turbine gearbox fault diagnosis method based on EMD-DCGAN method is proposed in this paper.&#x0D; OBJECTIVES: It can solve the problem when the sensor fails or the data transmission fails, it will lead to errors in the wind turbine gearbox fault data, which in turn will lead to a decrease in the wind turbine gearbox fault diagnosis accuracy.&#x0D; METHODS: Firstly, the outliers in the sample data need to be detected and removed. In this paper, the EMD method is used to eliminate outliers in the wind turbine gearbox fault data samples with the aim of enhancing the true continuity of the samples; secondly, in order to make up for the lack of missing samples, a data enhancement algorithm based on a GAN network is proposed in the paper, which is able to effectively perfect the missing items of the sample data; lastly, in order to improve the accuracy of wind turbine gearbox faults, a DCGAN neural network-based fault diagnosis method is proposed, which effectively combines the data dimensionality reduction feature of deep learning method and the data enhancement feature of generative adversarial network, and can improve the accuracy and speed of fault diagnosis.&#x0D; RESULTS and CONCLUSIONS: The experimental results show that the proposed method can effectively identify wind turbine gearbox fault conditions, and verify the effectiveness of the algorithm under different sample data conditions.

Thermodynamic and cloud evolution in a cold-air outbreak during HALO-(AC)3: quasi-Lagrangian observations compared to the ERA5 and CARRA reanalyses

Abstract. Arctic air masses undergo intense transformations when moving southward from closed sea ice to warmer open waters in marine cold-air outbreaks (CAOs). Due to the lack of measurements of diabatic heating and moisture uptake rates along CAO flows, studies often depend on atmospheric reanalysis output. However, the uncertainties connected to those datasets remain unclear. Here, we present height-resolved airborne observations of diabatic heating, moisture uptake, and cloud evolution measured in a quasi-Lagrangian manner. The investigated CAO was observed on 1 April 2022 during the HALO-(AC)3 campaign. Shortly after passing the sea-ice edge, maximum diabatic heating rates over 6 K h−1 and moisture uptake over 0.3 gkg-1h-1 were measured near the surface. Clouds started forming and vertical mixing within the deepening boundary layer intensified. The quasi-Lagrangian observations are compared with the fifth-generation global reanalysis (ERA5) and the Copernicus Arctic Regional Reanalysis (CARRA). Compared to these observations, the mean absolute errors of ERA5 versus CARRA data are 14 % higher for air temperature over sea ice (1.14 K versus 1.00 K) and 62 % higher for specific humidity over ice-free ocean (0.112 g kg−1 versus 0.069 g kg−1). We relate these differences to issues with the representation of the marginal ice zone and corresponding surface fluxes in ERA5, as well as the cloud scheme producing excess liquid-bearing, precipitating clouds, which causes a too-dry marine boundary layer. CARRA's high spatial resolution and demonstrated higher fidelity towards observations make it a promising candidate for further studies on Arctic air mass transformations.

Efficient Dynamic Distributed Resource Slicing in 6G Multi-Access Edge Computing Networks With Online ADMM and Message Passing Graph Neural Networks

We consider the problem of resource slicing in the 6 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">th</sup> generation multi-access edge computing (6G-MEC) network. The network includes many non-stationary space-air-ground-sea nodes with dynamic, unstable connections and resources, where any node can be in one of two hidden states: i) reliable – when the node generates/propagates no data errors; ii) unreliable – when the node can generate/propagate random errors. We show that solving this problem is challenging, since it represents a non-deterministic polynomial-time (NP) hard dynamic combinatorial optimization problem depending on the unknown distribution of hidden nodes' states and time-varying parameters (connections and resources of nodes) which can only be observed locally. To tackle these challenges, we develop a new deep learning (DL) model based on the message passing graph neural network (MPNN) to estimate hidden nodes' states in dynamic network environments. We then propose a novel algorithm based on the integration of MPNN-based DL and online alternating direction method of multipliers (ADMM) – extension of the well-known classical “static” ADMM to dynamic settings, where the slicing problem is solved distributedly, in real time, based on local information. We prove that our algorithm converges to a global optimum of our problem with a superior performance even in the highly-dynamic, unreliable scenarios.