This paper introduces a distributed edge intelligence framework for secure and efficient data sharing in 6G-Internet-of-Vehicles (6G-IoV) environments. The approach integrates a hierarchical blockchain architecture with an innovative asynchronous federated learning algorithm to address the challenges of privacy preservation, communication efficiency, and scalability in 6G-IoV scenarios. The proposed framework employs a multi-layer structure comprising vehicles, roadside units, and cloud servers, each playing a distinct role in the distributed learning process. We present a comprehensive latency model that accounts for computation and communication aspects, considering the unique characteristics of vehicular networks. The asynchronous federated learning algorithm incorporates genetic algorithm-based resource optimization to adaptively allocate communication resources, significantly enhancing efficiency in heterogeneous 6G-IoV environments. Furthermore, we introduce the hierarchical edge intelligence consensus mechanism, a lightweight consensus protocol tailored for edge intelligence scenarios, which accelerates blockchain consensus while maintaining security. Simulations using MNIST and SVHN datasets demonstrate that the proposed framework outperforms state-of-the-art methods regarding model accuracy, communication efficiency, and privacy preservation.

Consistency regularization and pseudo-labeling have significantly advanced semi-supervised learning (SSL). Prior works have effectively employed Mixup for consistency regularization in SSL. However, our findings indicate that applying Mixup for consistency regularization may degrade SSL performance by compromising the purity of artificial labels. Moreover, most pseudo-labeling based methods utilize thresholding strategy to exclude low-confidence data, aiming to mitigate confirmation bias; however, this approach limits the utility of unlabeled samples. To address these challenges, we propose RegMixMatch, a novel framework that optimizes the use of Mixup with both high- and low-confidence samples in SSL. First, we introduce semi-supervised RegMixup, which effectively addresses reduced artificial labels purity by using both mixed samples and clean samples for training. Second, we develop a class-aware Mixup technique that integrates information from the top-2 predicted classes into low-confidence samples and their artificial labels, reducing the confirmation bias associated with these samples and enhancing their effective utilization. Experimental results demonstrate that RegMixMatch achieves state-of-the-art performance across various SSL benchmarks.

Time series are often complex and rich in information but sparsely labeled and therefore challenging to model. Existing contrastive learning methods conduct augmentations and maximize their similarity. However, they ignore the similarity of adjacent timestamps and suffer from the problem of sampling bias. In this paper, we propose a self-supervised framework for learning generalizable representations of time series, called$\mathbf {R}$anking n$\mathbf {E}$ighborhood and cla$\mathbf {S}$s prototyp$\mathbf {E}$contr$\mathbf {A}$stive$\mathbf {L}$earning (RESEAL). It exploits information about similarity ranking to learn an embedding space, ensuring that positive samples are ranked according to their temporal order. Additionally, RESEAL introduces a class prototype contrastive learning module. It contrasts time series representations and their corresponding centroids as positives against truly negative pairs from different clusters, mitigating the sampling bias issue. Extensive experiments conducted on several multivariate and univariate time series tasks (i.e., classification, anomaly detection, and forecasting) demonstrate that our representation framework achieves significant improvement over existing baselines of self-supervised time series representation.

Accurate and robust time series forecasting is critical in numerous domains but remains challenging due to issues such as temporal misalignments, noise, and complex multivariate dependencies. Transformer-based models have demonstrated strong performance in sequential tasks; however, their reliance on dot-product attention renders them sensitive to noise and less effective for misaligned time series. To address these limitations, we propose a novel dynamic time warping (DTW)-based attention mechanism, leveraging a Sakoe-Chiba-constrained softDTW framework to replace the traditional dot-product similarity. This approach enables dynamic sequence alignment, enhancing robustness to temporal misalignments. Building on this innovation, we introduce DTWformer, a multi-scale Transformer model that integrates DTW-attention with adaptive patching to capture dependencies across varying temporal resolutions. DTWformer achieves superior forecasting performance and efficiency, addressing the limitations of existing approaches in handling misaligned and noisy time series. The implementation is publicly available at https://github.com/unihe/DTWformer.

The rapid growth of social media platforms has led to an unprecedented increase in user-generated content and social interactions, posing significant challenges for recommendation systems. This article addresses the challenges of recommendation in large-scale dynamic social environments, where user interactions and preferences evolve rapidly across vast networks. In large-scale dynamic social networks, knowledge discovery requires methods to efficiently process vast amounts of data while capturing the evolving nature of user interactions and preferences. Multigraph neural networks offer a promising approach for this task, as they can model complex relationships and temporal dynamics in these environments. This article proposes a novel social-aware multigraph neural network for the session-based recommendation (SAMGNN-SR) model that leverages dynamic social information and multigraph neural networks to enhance recommendation accuracy and knowledge discovery in complex social computing environments. The model constructs a global social-aware interaction graph from all user session sequences and employs an adaptive subgraph sampling strategy to extract relevant collaborative signals efficiently. A dynamic interest extraction module utilizing dual-direction information propagation captures users' evolving preferences, while a social information fusion network based on graph attention mechanisms models the dynamic nature of social influences. Experiments on three real-world datasets (Douban, Delicious, and Yelp) demonstrate the superiority of SAMGNN-SR over nine state-of-the-art baselines, with improvements of up to 6.79% in NDCG@20 and 6.19% in Hit@20. Ablation studies validate the effectiveness of each model component in capturing complex social dynamics and session-based user behaviors.

Considering the extremely low signal-to-noise ratio (SNR) environment in future satellite internet of Things (IoT) transmission in which the traditional Hamming coded LoRa (Long Range) system in the ground IoT communication is of difficulty to guarantee high reliability, optimal coding scheme selection for the LoRa-based satellite IoT is studied for the first time. In specific, the corresponding system model is described in detail, followed by derivations of two noncoherent detection methods of coded LoRa signals. Subsequently, the channel capacity of the coded LoRa system is provided and used to develop the corresponding relationship between the signal-to-noise ratio (SNR) and the code rate. Based on the obtained relationship and the Monte-Carlo method, it is found that the coded LoRa system with different spreading factors (SFs) can achieve the lowest SNR when the code rate is close to 1/2. Finally, the performance of several coded LoRa systems (including Hamming code, BCH code, RS code, Turbo code, LDPC code and LDGM code with code rates of about 1/2) is compared by means of the Monte-Carlo method. Simulation results show that Turbo or LDPC coded LoRa systems exhibit more than 4.5 dB performance advantages over the Hamming coded LoRa system.

Considering limited pilot overhead and large delay and Doppler shift in LoRa (Long Range)-based IoT (Internet of Things) systems, this paper presents an efficient joint parameter estimation and soft nonconherent detection scheme. During the parameter estimation, a joint delay and Doppler shift estimation method is proposed, which is called the JDDSE method. Different from the existing estimation methods, the proposed method utilizes the time-domain characteristic of the LoRa signal and only needs to estimate a joint offset of the delay and Doppler shift instead of their respective offsets. Then, a soft noncoherent detection is developed from the existing work in order to match the JDDSE method. Simulation results indicate the validity of the proposed method under the uplink IoT transmission.

A novel low-density parity-check (LDPC) coded Gaussian minimum shift keying (GMSK) scheme is proposed for wireless communications subject to low SNRs, limited power and spectrum resources. We first design a non-recursive GMSK modulator to alleviate the impact of error propagation. Then, a pilot-aided quasi-coherent demodulation algorithm (PA-QCDA) is derived, where a modified BCJR-based detection is used to produce the soft-output with initial and ending trellis-states being determined using the overhead-limited pilot. We choose proper parameters for the non-recursive GMSK signaling according to the trade-off of the power and spectral efficiency. Simulation results show that the proposed non-recursive GMSK system with the PA-QCDA can achieve performance similar to the LDPC coded BPSK system and can also work well in the presence of large frequency and phase offsets or burst errors.

In this paper, we consider the sparse code multiple access (SCMA) for low earth orbit (LEO) satellite communications to achieve high user capacity and provide technical support for satellite-based Internet of Things (IoT) applications. Note that Doppler shift in the LEO satellite link can not be ignored and the message passing algorithm (MPA) extensively utilized by SCMA is quite time-consuming due to high complexity. Thus we propose a coherent iterative receiver with low complexity for SCMA-based LEO satellite communication systems. First, we design a data-aided carrier synchronization scheme using multiple disjoint pilot blocks to estimate the Doppler shift precisely. Then, a soft-out approximate message passing (SO-AMP) algorithm with linear complexity is derived from the first-order approximation of MPA based on a novel framework of SCMA mapper. Further, a damping method is introduced into the SO-AMP detector developing a damped SO-AMP (DSO- AMP) to improve the overall performance. Numerical results show that the designed receiver can achieve excellent performance close to ideal performance with no Doppler shift while maintaining low complexity.

In this paper, we design a generalized pilot symbol assisted modulation (G-PSAM) format and study its relevant optimization problem to serve unsynchronized wireless communication systems with small pilot overhead. Based on the prototype of the PSAM format, we design a generalized PSAM format and derive corresponding data-aided Cramer-Rao bound (DA CRB) and data-aided & non-data-aided CRB (DA&NDA CRB). Then taking the DA CRB for frequency offset estimation as a criterion and adopting control variable method, we solve GPSAM format optimization problem and rebuild two “good” GPSAM formats. Theoretical analysis and simulation results show that the two G-PSAM formats can achieve better performance than the existing special PSAM formats (including the classical preamble-postamble (PP) format); and an unbiased frequency offset estimator using the two G-PSAM formats will tend more possibly to employ easier DA processing rather than DA&NDA processing without significant performance loss.

In this paper, we consider the detection of faster-than-Nyquist (FTN) signaling. By noticing that the whitening filter for FTN signaling cannot be directly derived when the symbol rate exceeds the signal bandwidth, we propose a new reduced-complexity M-algorithm BCJR (M-BCJR) algorithm based on the Ungerboeck observation model. By taking some “future” symbols into account, the proposed algorithm is able to select the M best states in the maximum a posteriori sense. We further simplify the above algorithm by choosing the key path from each possible state, which successfully reduces the complexity while maintaining a good bit error rate performance. Simulation results show that, with the use of the proposed methods, great gains can be obtained in terms of spectral efficiency (up to 186%) or signal-to-noise ratio (up to 4.5 dB) compared with the Nyquist signaling.

In this paper, a two-stage phase-increment-based frequency estimator with a general pilot-symbol-assisted- modulation (PSAM) structure is proposed for burst-mode communications. We derive the true Cramer-Rao bounds (CRBs) under the general PSAM structure for pure data-aided (DA) estimator. The proposed phase increment-based estimator is a pure DA estimator and implemented via two stages. In the first stage, a joint auto-correlation with one-symbol delay (AC-1) and simplified L&R (S-L&R) estimator is employed for the coarse estimation. In another stage, a two-step simplified cross-correlation algorithm with three disjoint pilot blocks is proposed for the fine estimation, where the problem of phase ambiguity is solved in the first stage such that wide estimation range as large as half of symbol rate can be guaranteed, and the fine estimation improves overall estimation accuracy. It is shown that the proposed estimator exhibits wide operating range while providing both high accuracy and moderate complexity, and efficient pilot utilization.