Stereo image sand removal is crucial to improve the perceptual quality for autonomous driving perception. Existing methods often fall short in accurately estimating the uncertainty inherent in degraded images, leading to suboptimal outcomes. To address this, we introduce a novel framework named Decoupling While Coupling(DWC). DWC pioneers the integration of inter-view uncertainty estimation, cross-view uncertainty-aware interaction and block-wise uncertainty representation for superior stereo image sand removal. For cross-view information interaction, we propose an Uncertainty-aware Cross-view Attentive Interaction module(UCAI) to cope with the lack of uncertainty estimation ability in the existing cross-view information interaction mechanism. For the uncertainty perception and information interaction within the inter-view, we propose a Distribution Modeling Coupling Block(DMCB), which transmits the representation of uncertainty between each backbone module. For block-wise uncertainty estimation, we use our proposed Uncertainty-aware Distribution Feature Modulator(UDFM) as the backbone of DWC to modulate the uncertainty inside the neural network itself. Extensive experimental validations on our proposed stereo image sand removal dataset SandST confirm the efficacy of DWC. Our method not only achieves higher PSNR and SSIM, but also exhibits enhanced robustness against various sand degrees and patterns.

The growing diversity of 5G services with stringent service-level agreement (SLA) requirements has created an urgent need for efficient radio access network (RAN) slicing solutions. This paper investigates a cell-free architecture for SLA-guaranteed RAN slicing in densely populated scenarios, addressing the limitations of conventional small-cell architectures through enhanced spatial diversity and distributed coordination. We propose a twin-timescale resource allocation framework. At large timescales, we propose a distributed generative adversarial network (GAN)-based deep reinforcement learning (DRL) framework using the innovative reward clipping mechanism to maximize system utility, which is defined as a weighted combination of the SLA satisfaction ratio (SSR) and spectral efficiency (SE). At small timescales, we apply offline beam fingerprinting and online fast selection to determine the precoding vector and use slice-aware schedulers to allocate RBs to the user level. Simulation results show that the beamforming scheme has achieved near-optimal performance and validated that our cell-free architecture achieves superior system utility and SLA robustness compared to small-cell baselines, especially with high user density.

As a key technology in beyond 5G and future 6G communications, RAN slicing can realize differentiated service level agreement (SLA) guarantees. In this paper, we investigate the multi-slice multi-user RAN slicing. A dynamic bandwidth allocation scheme for RAN slicing is proposed based on hierarchical intelligence, where bandwidth pre-allocation and hyper-parameter tuning for MAC layer schedulers are jointly optimized to maximize the system utility, i.e., the weighted sum of spectrum efficiency (SE) and SLA satisfaction ratio (SSR) of different slices. The problem is formulated as a twin-time scale Markov decision process (MDP), where the bandwidth pre-allocation and the scheduler parameter tuning are performed on a long-term scale (e.g., seconds) and a short-term scale (e.g., 100 ms), respectively, for which we propose a hierarchical twin-time scale Dueling deep Q learning network (TTS-DDQN) algorithm. A new reward-clipping mechanism is proposed to get a better trade-off between stabilized training and higher system utility. In order to improve the robustness to time-varying traffic patterns and non-stationary dynamic environments, we further propose a traffic-aware module for more efficient sampling of the experience pool, and a variational adversarial inverse reinforcement learning (VAIRL) module for reward automation design. Extensive simulations show that the traffic-aware TTS-DDQN in stationary scenarios and the VAIRL module embedded TTS-DDQN in non-stationary scenarios outperform existing typical DQN-based algorithms, hard slicing and non-slicing, etc.

Object detection in UAV remote sensing imagery is significantly challenged by small-scale objects, dense distributions, and complex backgrounds. Existing DETR-based models struggle with multi-scale feature extraction, and traditional multi-head self-attention (MHSA) often introduces redundancy and noise when handling high-frequency details. Moreover, current methods lack effective multi-domain feature modeling. To address these issues, this paper proposes a Scale-Aware Multi-Domain DETR (SME-DETR) for small object detection in UAV remote sensing imagery. Firstly, a Scale-Aware Attentional Backbone (SAA-Backbone) is designed, where the Dynamic Scale-Aware module (DSA) adaptively fuses multi-scale features using depthwise separable convolutions and dynamic weighting. In addition, a Directional Channel Attention Module (DCA) further enhances edge and texture representation. Then, a Multi-domain Augmented Pyramid (MDAP) is constructed, integrating a CSP-based Multi-domain Kernel (CSP-MDKernel) to jointly optimize spatial, frequency, and channel domain features. A Spatial-Channel Fusion Convolution (SCFConv) is employed to preserve fine-grained details. Finally, an Efficient Attention-based Intra-scale Feature Interaction module (EfficientAIFI) is proposed, which replaces the complex matrix operations of traditional multi-head self-attention (MHSA) with linear element-wise multiplication. At the same time, it maintains global dependencies through normalized inner products between queries and keys. Additionally, a lightweight Background-Suppression Gate (BSG) is incorporated to mitigate false positives induced by cluttered backgrounds, further enhancing detection robustness. Experiments on the VisDrone-DET, RSOD, and SeaDronesSeeV2 datasets with the core SME-DETR framework demonstrate that SME-DETR achieves [email protected] scores of 53.0% and 97.8% on the VisDrone-DET and RSOD, respectively, improving the RT-DETR-r18 baseline by 6.0% and 3.0%, respectively. Moreover, SME-DETR significantly outperforms most state-of-the-art detectors.

In this paper, the cross-layer optimization problem of access point selection (APS) and beamforming (BF) in cell-free network (CFN) with local CSI is studied, where constraints of per AP power and the number of active APs are considered. Such a joint APS&BF optimization problem is modeled as a mixed-integer nonlinear programming (MINP) problem aiming at maximizing the sum rate of the whole system. Fractional programming (FP)-based and alternating optimization (AO)-based algorithms with weightedl1-norm approximation are proposed to solve this MINP problem. However, the latter performs better than the former, with higher complexity. A lightweight multi-head single-body graph neural network (MHSB-GNN) algorithm is proposed, where the nodes and structures are innovatively designed. The MHSB-GNN benefits from the different node updating modules for different user equipment (UE), which introduce extra prior information into the graph and mine specific information of different UEs. Moreover, the equivalence between GNN and FP-based algorithm is proved to provide interpretability and theoretical guarantees for MHSB-GNN. The analysis of convergence and complexity validates the accuracy and effectiveness of the FP and AO-based algorithms. Leveraging the existing APS and BF solver, it is shown that the three proposed algorithms guarantee comparable performance as the exhaustive search algorithm in performance and complexity.

Despite profound impacts on human health and nature, accurately predicting microbe-metabolite interactions remains challenging due to inherent data noise. This study applies non-negative matrix factorization (NMF) and multi-view NMF to reduce noise and exploit associations across data perspectives. NMF obtains low-dimensional microbial and metabolic representations, effectively reducing noise. The dimension-reduced spectral matrices were input into the generative network model to derive conditional probabilities of individual microbe-associated metabolites and microbe-metabolite co-occurrence probabilities, the latter enabling prediction of microbe-metabolite interactions. Moreover, multi-view NMF integrates microbial and metabolic data by mapping them into a shared subspace, thereby enhancing prediction performance and validating cross-perspective correlation modeling. This study demonstrates NMF's efficacy in noise reduction through dimensionality reduction, and multiview NMF's ability to leverage cross-view associations. Both approaches demonstrate enhanced microbe-metabolite interaction prediction utilizing NMF-based and multi-view NMF-based generative network models.

RAN slicing is a key technology in 5G communications for realizing differentiated service level agreement (SLA) guarantee. In this paper, we investigate the multi-slice multi-user RAN slicing. A dynamic bandwidth allocation scheme for RAN slicing is proposed based on hierarchical intelligence, where bandwidth pre-allocation and hyper-parameter tuning for MAC-layer schedulers are jointly optimized to maximize the system utility, i.e., the weighted sum of spectrum efficiency (SE) and SLA satisfaction ratio (SSR) of different slices. The problem is formulated as a twin-time scale Markov decision process (MDP), where the bandwidth pre-allocation and the scheduler parameter tuning are performed on a long-term scale (e.g., seconds) and a short-term scale (e.g., 100 ms), respectively, for which we propose a hierarchical twin-time scale Dueling deep Q learning network (TTS-DDQN) algorithm. And a new reward-clipping mechanism is proposed to get a better trade-off between stabilized training and higher system utility. In order to improve the algorithm robustness to time-varying traffic pattern, we further propose a traffic-aware module for more efficient sampling of the experience pool. Extensive simulations show that the traffic-aware TTS-DDQN outperforms existing typical DQN based algorithms, hard slicing and non slicing, etc.

In this paper, a cross-layer optimization problem of access point selection (APS) and beamforming (BF) in cell free network (CFN) has been studied, where constraints of per access point (AP) power and per user receiving data streams are considered. Such a cross-layer design of APS&BF problem is modeled as a mixed-integer nonlinear programming (MINP) program. Then, by adopting the weighted l1-norm approximation, the MINP problem is transformed into the sum logarithmic multiple-ratio form. To be specific, a novel and low-complexity mix-integer fractional programming (MIFP) algorithm is proposed to solve the transformed problem effectively. Convergence analysis validates that the proposed MIFP converges to a local optimal solution. Finally, numerical results show that cross-layer design of APS&BF scheme is superior to separate design of APS&BF schemes. In addition, the proposed MIFP has the approximate performance as partial exhaustive search algorithm.

Mobile-Edge Computing (MEC) is an emerging paradigm in the Internet of Vehicles (IoV) to meet the ever-increasing computation demands of smart applications. To provide satisfactory computation performance, it is of significant importance to conduct computation offloading in IoV. In this paper, we investigate a multi-vehicle IoV system assisted by MECs with limited computation resources, where vehicles with complex applications can offload their subtasks to MEC servers. Applications are modeled as interdependent subtasks with general random task graphs, different from existing works with independent ones. To maximize the average logarithmic data processing rate (LDPR), the computation offloading problem is formulated as a time-average optimization with long-term constraints, which results from variable vehicle number, various applications and time-varying communication channels. To reduce the cooperation overhead, we propose a multi-agent Proximal Policy Optimization algorithm (Ly-MAPPO) which requires local observation only to solve the subproblems achieved by Lyapunov optimization technique in real time. In addition, to improve the performance of the Ly-MAPPO algorithm, Graph Convolutional Neural Network (GCN) is introduced to extract inter-dependencies between subtasks. Extensive simulations show that the GCN embedded Ly-MAPPO outperforms other baseline algorithms, e.g., greedy algorithm and gene algorithm, etc., for different traffic loads and computation resources in MEC servers.

Approximate computing is a new design paradigm targeting at error-tolerant applications. By allowing a little amount of inaccuracy in the computation, it could significantly reduce circuit area and power consumption. Several logic synthesis methods for approximate computing were proposed recently. However, these methods are mainly aimed at ASIC designs. In this work, we propose a novel approximate logic synthesis method targeting at the FPGA design. We exploit the flexibility of lookup tables and propose a method that combines wire removal and local function change. The experimental results showed that our method produces better results than the state-of-the-art approximate logic synthesis method adapted to FPGA designs. Moreover, it can be combined with the state-of-the-art method to further improve the design quality.

Non-stationary Dynamic Bayesian Networks (Non-stationary DBNs) are widely used to model the temporal changes of directed dependency structures from multivariate time series data. However, the existing change-points based non-stationary DBNs methods have several drawbacks including excessive computational cost, and low convergence speed. In this paper we proposed a novel non-stationary DBNs method. Our method is based on the perfect simulation model. We applied this approach for network structure inference from synthetic data and biological microarray gene expression data and compared it with other two state-of-the-art non-stationary DBNs methods. The experimental results demonstrated that our method outperformed two other state-of-the-art methods in both computational cost and structure prediction accuracy. The further sensitivity analysis showed that once converged our model is robust to large parameter ranges, which reduces the uncertainty of the model behavior.

BACKGROUND: Dynamic Bayesian Networks (DBNs) are widely used in regulatory network structure inference with gene expression data. Current methods assumed that the underlying stochastic processes that generate the gene expression data are stationary. The assumption is not realistic in certain applications where the intrinsic regulatory networks are subject to changes for adapting to internal or external stimuli. RESULTS: In this paper we investigate a novel non-stationary DBNs method with a potential regulator detection technique and a flexible lag choosing mechanism. We apply the approach for the gene regulatory network inference on three non-stationary time series data. For the Macrophages and Arabidopsis data sets with the reference networks, our method shows better network structure prediction accuracy. For the Drosophila data set, our approach converges faster and shows a better prediction accuracy on transition times. In addition, our reconstructed regulatory networks on the Drosophila data not only share a lot of similarities with the predictions of the work of other researchers but also provide many new structural information for further investigation. CONCLUSIONS: Compared with recent proposed non-stationary DBNs methods, our approach has better structure prediction accuracy By detecting potential regulators, our method reduces the size of the search space, hence may speed up the convergence of MCMC sampling.

Graph data mining is an active research area. Graphs are general modeling tools to organize information from heterogeneous sources and have been applied in many scientific, engineering, and business fields. With the fast accumulation of graph data, building highly accurate predictive models for graph data emerges as a new challenge that has not been fully explored in the data mining community. In this paper, we demonstrate a novel technique called graph pattern diffusion (GPD) kernel. Our idea is to leverage existing frequent pattern discovery methods and to explore the application of kernel classifier (e.g., support vector machine) in building highly accurate graph classification. In our method, we first identify all frequent patterns from a graph database. We then map subgraphs to graphs in the graph database and use a process we call "pattern diffusion" to label nodes in the graphs. Finally, we designed a graph alignment algorithm to compute the inner product of two graphs. We have tested our algorithm using a number of chemical structure data. The experimental results demonstrate that our method is significantly better than competing methods such as those kernel functions based on paths, cycles, and subgraphs.

Dynamic Bayesian networks (DBNs) are widely used in regulatory network structure inference from noisy gene expression data. However most of the previous researches assumed that the underlying stochastic processes that generates the gene expression data are stationary. Such assumption is not realistic in certain applications where the intrinsic regulatory networks are subject to change for adapting to internal or external stimuli. In this paper we investigate a novel non-stationary DBNs method and apply the approach for the gene regulatory network inference on Arabidopsis thaliana circadian time series data. Our experimental study demonstrated that compared with recent proposed non-stationary DBNs methods, our approach has better structural prediction performance, and can potentially reduce the computational cost by improving the sampling convergence speed.

BACKGROUND: Automatic identification of structure fingerprints from a group of diverse protein structures is challenging, especially for proteins whose divergent amino acid sequences may fall into the "twilight-" or "midnight-" zones where pair-wise sequence identities to known sequences fall below 25% and sequence-based functional annotations often fail. RESULTS: Here we report a novel graph database mining method and demonstrate its application to protein structure pattern identification and structure classification. The biologic motivation of our study is to recognize common structure patterns in "immunoevasins", proteins mediating virus evasion of host immune defense. Our experimental study, using both viral and non-viral proteins, demonstrates the efficiency and efficacy of the proposed method. CONCLUSION: We present a theoretic framework, offer a practical software implementation for incorporating prior domain knowledge, such as substitution matrices as studied here, and devise an efficient algorithm to identify approximate matched frequent subgraphs. By doing so, we significantly expanded the analytical power of sophisticated data mining algorithms in dealing with large volume of complicated and noisy protein structure data. And without loss of generality, choice of appropriate compatibility matrices allows our method to be easily employed in domains where subgraph labels have some uncertainty.

BACKGROUND: A primary reason for using two-color microarrays is that the use of two samples labeled with different dyes on the same slide, that bind to probes on the same spot, is supposed to adjust for many factors that introduce noise and errors into the analysis. Most users assume that any differences between the dyes can be adjusted out by standard methods of normalization, so that measures such as log ratios on the same slide are reliable measures of comparative expression. However, even after the normalization, there are still probe specific dye and slide variation among the data. We define a method to quantify the amount of the dye-by-probe and slide-by-probe interaction. This serves as a diagnostic, both visual and numeric, of the existence of probe-specific dye bias. We show how this improved the performance of two-color array analysis for arrays for genomic analysis of biological samples ranging from rice to human tissue. RESULTS: We develop a procedure for quantifying the extent of probe-specific dye and slide bias in two-color microarrays. The primary output is a graphical diagnostic of the extent of the bias which called ECDF (Empirical Cumulative Distribution Function), though numerical results are also obtained. CONCLUSION: We show that the dye and slide biases were high for human and rice genomic arrays in two gene expression facilities, even after the standard intensity-based normalization, and describe how this diagnostic allowed the problems causing the probe-specific bias to be addressed, and resulted in important improvements in performance. The R package LMGene which contains the method described in this paper has been available to download from Bioconductor.