Journal of King Saud University-Computer and Information Sciences最新文献

Heterogeneous network link prediction is a hot topic in the analysis of networks. It aims to predict missing links in the network by utilizing the rich semantic information present in the heterogeneous network, thereby enhancing the effectiveness of relevant data mining tasks. Existing heterogeneous network link prediction methods utilize meta-paths or meta-graphs to extract semantic information, heavily relying on the priori knowledge. This paper proposes a heterogeneous network link prediction based on network schema and cross-neighborhood attention method (HNLP-NSCA). The heterogeneous node features are projected into a shared latent vector space using fully connected layers. To resolve the issue of prior knowledge dependence on meta-path, the semantic information is extracted by using network schema structures uniquely in heterogeneous networks. Node features are extracted based on the relevant network schema instances, avoiding the problem of meta-path selection. The neighborhood interaction information of input node pairs is sensed via cross-neighborhood attention, strengthening the nonlinear mapping capability of the link prediction. The resulting cross-neighborhood interaction vectors are combined with the node feature vectors and fed into a multilayer perceptron for link prediction. Experimental results on four real-world datasets demonstrate that the proposed HNLP-NSCA mothed outperforms the baseline models.

Transformer-based approaches have demonstrated remarkable performance in image processing tasks due to their ability to model long-range dependencies. Current mainstream Transformer-based methods typically confine self-attention computation within windows to reduce computational burden. However, this constraint may lead to grid artifacts in the reconstructed images due to insufficient cross-window information exchange, particularly in image super-resolution tasks. To address this issue, we propose the Multi-Scale Texture Complementation Block based on Spatial Relaxation Transformer (MSRT), which leverages features at multiple scales and augments information exchange through cross windows attention computation. In addition, we introduce a loss function based on the prior of texture smoothness transformation, which utilizes the continuity of textures between patches to constrain the generation of more coherent texture information in the reconstructed images. Specifically, we employ learnable compressive sensing technology to extract shallow features from images, preserving image features while reducing feature dimensions and improving computational efficiency. Extensive experiments conducted on multiple benchmark datasets demonstrate that our method outperforms previous state-of-the-art approaches in both qualitative and quantitative evaluations.

The rapidly-exploring random tree (RRT) algorithm has a wide range of applications in the field of path planning. However, conventional RRT algorithm suffers from low planning efficiency and long path length, making it unable to handle complex environments. In response to the above problems, this paper proposes an improved RRT algorithm based on restricted sampling area (RSA-RRT). Firstly, to address the problem of low efficiency, a restricted sampling area strategy is proposed. By dynamically restricting the sampling area, the number of invalid sampling points is reduced, thus improving planning efficiency. Then, for the path planning problem in narrow areas, a fixed-angle sampling strategy is proposed, which improves the planning efficiency in narrow areas by conducting larger step size sampling with a fixed angle. Finally, a multi-triangle optimization strategy is proposed to address the problem of longer and more tortuous paths. The effectiveness of RSA-RRT algorithm is verified through improved strategy performance verification and ablation experiments. Comparing with other algorithms in different environments, the results show that RSA-RRT algorithm can obtain shorter paths while taking less time, effectively balancing the path quality and planning speed, and it can be applied in complex real-world environments.

Reversible data hiding (RDH) algorithms have been extensively employed in the field of copyright protection and information dissemination. Among various RDH algorithms, the multiple histogram modification (MHM) algorithm has attracted significant attention because of its capability to generate high-quality marked images. In previous MHM methods, the prediction error histograms were mostly generated in a fixed way. Recently, clustering algorithms automatically classify prediction errors into multiple classes, which enhances the similarity among prediction errors within the same class. However, the design of features and the determination of clustering numbers are crucial in clustering algorithms. Traditional algorithms utilize the same features and fix the number of clusters (e.g., empirically generate 16 classes), which may limit the performance due to the lack of adaptivity. To address these limitations, an adaptive initial feature selection scheme and a clustering number optimization scheme based on the Fuzzy C-Means (FCM) clustering algorithm are proposed in this paper. The superiority of the proposed scheme over other state-of-the-art schemes is verified by experimental results.

Due to its asymptotic optimality, the Rapidly-exploring Random Tree star (RRT*) algorithm is widely used for robotic operations in complex environments. However, the RRT* algorithm suffers from poor path quality, slow convergence, and unstable generation of high-quality paths in the path planning process. This paper proposes an Improved Bi-Tree Obstacle Edge Search Artificial Potential Field RRT* algorithm (IBPF-RRT*) to address these issues. First, based on the RRT* algorithm, this paper proposes a new obstacle edge search artificial potential field strategy (ESAPF), which speeds up the path search and improves the path quality simultaneously. Second, a bi-directional pruning strategy is designed to optimize the bi-directional search tree branch nodes and combine the bi-directional search strategy to reduce the number of iterations for convergence speed significantly. Third, a novel path optimization strategy is proposed, which enables high-quality paths to be generated stably by creating an entirely new node between two path nodes and then optimizing the paths using a pruning strategy based on triangular inequalities. Experimental results in three different scenarios show that the proposed IBPF-RRT* algorithm outperforms other methods in terms of optimal path quality, algorithm stability, and the number of iterations when compared to the RRT*, Q-RRT*, PQ-RRT*, F-RRT* and CCPF-RRT* algorithms, and proves the effectiveness of the proposed three strategies.

In a service ecosystem, the trust of users in services serves as the foundation for maintaining normal interactions among users, service providers, and platforms. However, malicious attacks can tamper with the trust value of these services, making it difficult for users to identify reliable services and undermining the benefits of reliable service providers and platforms. When existing trust management models address the impact of malicious attacks on service reliability, they rarely consider leveraging different attack targets to improve the accuracy of compromised service trust. Therefore, we propose a trust enhancement model based on distributed learning and blockchain in the service ecosystem, which adaptively enhances the trust values of compromised services according to the targets of anomalous attacks. Firstly, we conduct a comprehensive analysis of the targets of malicious attacks using distributed learning. Secondly, we introduced a trust enhancement contract that utilizes different methods to enhance the trust of the service based on various attack targets. Finally, our approach outperforms the baseline method significantly. For different attack targets, we observe a reduction in RMSE by 12.38% and 12.12%, respectively, and an enhancement in coverage by 24.94% and 14.56%, respectively. The experimental results show the reliability and efficacy of our proposed model.

Low back pain (LBP) is a prevalent spinal issue, affecting eight out of ten individuals. Notably, lumbar intervertebral disc (IVD) abnormalities frequently contribute to LBP. To diagnose LBP, Magnetic Resonance Imaging (MRI) is crucial for obtaining detailed spinal images. This paper employs deep learning (DL) to detect and locate lumbar IVD in sagittal MR images. It further classifies lumbar IVDs as healthy or herniated, utilizing both novel convolutional neural network (CNN) and conventional CNN models. The dataset utilized comprises MR images from 32 patients, with 10 exhibiting healthy discs and the remaining 22 posing a mix of healthy and herniated discs, totaling 160 lumbar discs, incorporating 112 healthy and 48 herniated discs. In this study, ResNet-50 architecture in the Novel Lumbar IVD detection (NLID) model served as the feature extractor to segment the five lumbar IVDs from MR images. The features extracted from ResNet-50 were input into YOLOv2 for the identification of the region of interest (ROI). The findings indicate that optimal performance was achieved at the 22nd Rectified Linear Unit (ReLU) activation layer, boasting a remarkable 99.59% average precision, 97.22% F1-score, 94.59% precision, and a perfect 100% recall. This commendable performance consistently held above the 85% threshold until the 22nd ReLU activation layer. Regarding imbalanced dataset classification, AlexNet emerged as the frontrunner among other pre-trained networks, boasting the highest test accuracy of 90.63%, and an impressive F1 score of 88.77%. Meanwhile, the Novel Lumbar IVD Classification (NLIC) model achieved superior results with 93.75% test accuracy, and 92.27% F1-score. In the setting of the balanced dataset, NLIC achieved 96.88% test accuracy, and 96.46% F1-score with fewer epochs compared to AlexNet, affirming the robustness of the novel trained-from-scratch network. These findings distinctly underscore the effectiveness of CNNs in both medical image segmentation and classification.

Synthesis of Text-to-speech (TTS) is a process that involves translating a natural language text into a speech. Speech synthesisers face a major challenge when recognizing the prosodic elements of written text, such as intonation (the rise and fall of the voice in speaking), and length. In contrast, continuous speech features are influenced by the personality and emotions of the artist. A database is maintained to store the synthesized speech pieces. Its output is determined by how similar the person utters the words and how capable they are of being implied. In the past few years, the field of text-to-speech synthesis has been heavily impacted by the emergence of deep learning, an AI technology that has gained widespread popularity. This review paper presents a taxonomy of models and architectures that are based on deep learning and discusses the various datasets that are utilised in the TTS process. It also covers the evaluation matrices that are commonly used. The paper ends with a look at the future directions of the system and reaches to some Deep learning models that give promising results in this field.

Weeds in rapeseed field are an important factor in crop yield reduction and economic loss. Thus, Precision Agriculture is an important task for sustainable agriculture and weed management. At present, deep learning techniques have shown great potential for image-based detection and classification in various crops and weeds. However, the inherent limitations of traditional convolutional neural networks pose significant challenges due to the locally similarity of weeds and crops in color, shape and texture. To address this issue, we introduce SWFormer, a scale-wise hybrid CNN-Transformer network. SWFormer leverages the distinct strengths of both convolutional and transformer architectures. Convolutional structures excel at extracting short-range dependency information among pixels, whereas transformer structures are adept at capturing global dependency relationships. Additionally, we propose two innovative modules. Firstly, the Scale-wise Cascade Convolution (SWCC) module is designed to capture multiscale features and expand the receptive field. Secondly, the Adaptive Semantic Aggregation (ASA) module facilitates adaptive and effective information fusion across two distinct feature maps. Our experiments were conducted on the publicly available cropandweed dataset and SB20 dataset. it yields improved performance over other mainstream segmentation models. Specifically, SWFormer with 52.33M/527.51GFLOPs achieves an mAP of 76.54% and an accuracy of 83.95% on the cropandweed dataset. For the SB20 dataset, it attains an mAP of 61.24% and an accuracy of 79.47%. Overall, the evaluation clearly demonstrates our proposed SWFormer is conducive to promoting further research in the area of Precision Agriculture.

Multi-view graph clustering has garnered tremendous interest for its capability to effectively segregate data by harnessing information from multiple graphs representing distinct views. Despite the advances, conventional methods commonly construct similarity graphs straightway from raw features, leading to suboptimal outcomes due to noise or outliers. To address this, latent representation-based graph clustering has emerged. However, it often hypothesizes that multiple views share a fixed-dimensional coefficient matrix, potentially resulting in useful information loss and limited representation capabilities. Additionally, many methods exploit Euclidean distance as a similarity metric, which may inaccurately measure linear relationships between samples. To tackle these challenges, we develop a novel diverse representation-guided graph learning for multi-view metric clustering (DRGMMC). Concretely, raw sample matrix from each view is first projected into diverse latent spaces to capture comprehensive knowledge. Subsequently, a popular metric is leveraged to adaptively learn similarity graphs with linearity-aware based on attained coefficient matrices. Furthermore, a self-weighted fusion strategy and Laplacian rank constraint are introduced to output clustering results directly. Consequently, our model merges diverse representation learning, metric learning, consensus graph learning, and data clustering into a joint model, reinforcing each other for holistic optimization. Substantial experimental findings substantiate that DRGMMC outperforms most advanced graph clustering techniques.