Social relations are often used as auxiliary information to improve recommendations. In the real-world, social among users complex and diverse. However, most existing recommendation methods assume only single relation (i.e., exploit pairwise mine user preferences), ignoring impact of multifaceted on preferences high order complexity relations). Moreover, an observing fact is that similar items always have attractiveness when exposed users, indicating a potential connection static attributes...

Automatic Grammar Error Correction (GEC) detects and corrects various types of syntax, spelling, grammatical errors. Different approaches such as rule-based, Statistical Machine Translation (SMT), Neural (NMT) have been proposed. Among these approaches, NMT based on seq2seq multi-head attention (Transformer) performs the best. The key shortcoming GEC models with multiple encoder-decoder layers is that only top layer exploited in subsequent processes. In addition, due to exposure bias problem...

Big data is promoting the development of supply chain design and management. The problem trustworthy scheduling by using big challenging, it significantly influences performance agricultural products (APSC) Currently, there are various approaches to optimize APSC, but most them can only tackle with primary objectives (time cost) or limited small-scale chains. efficient have not been provided for APSC in environment. This paper aims at proposing a novel optimization approach data. First, new...

Social relations are often used as auxiliary information to address data sparsity and cold-start issues in social recommendations. In the real world, among users complex diverse. Widely graph neural networks (GNNs) can only model pairwise node relationships not conducive exploring higher-order connectivity, while hypergraph provides a natural way high-order between nodes. However, recent studies show that recommendations still face following challenges: 1) majority of ignore impact...

Deep learning network models are crucial in processing images acquired from optical, laser, and acoustic sensors ocean intelligent perception target detection. This work comprehensively reviews image technology, including devices acquisition, recognition detection models, adaptive processes, coping methods for nonlinear noise interference. As the core tasks of processing, research focus this article. The is on development deep-learning detection, such as SSD, R-CNN series, YOLO series....

Astounding results from transformer models with Vision-and Language Pretraining (VLP) on joint vision-and-language downstream tasks have intrigued the multi-modal community. On one hand, these are usually so huge that make us more difficult to fine-tune and serve real-time online applications. other compression of original block will ignore difference in information between modalities, which leads sharp decline retrieval accuracy.

Cross-lingual word alignment is the task for translation between monolingual embedding spaces of two different languages. Recent work mostly based on supervised approaches, while their success relies bilingual seed dictionaries derived from aligned data. The unsupervised adversarial which utilize generative networks (GANs) to map global space another space, can eliminate need However, most GAN-based approaches ignore issues mode collapse and gradient disappearance in GANs, leading a training...

Wafer defect detection is a critical process in semi-conductor manufacturing, ensuring the quality and reliability of chip production. In this study, wafer network inspired by YOLOv8 was proposed. The algorithm integrates two new features, C2f Spatial Pyramid Pooling Fusion (SPPF), to enhance accuracy model. module improves model's feature representation ability, while SPPF captures multi-scale features from input images. Experimental results on common dataset WM-811K demonstrate that...

A clustering algorithm combining particle swarm optimization (CPSO) with K-Means (KM-CPSO) is proposed, which features better search efficiency than K-Means, PSO and CPSO.The algorithms cannot guarantee convergence to global optima suffer in local optimal cluster center because they are sensitive initial centers.Chaotic can find solution; meanwhile achieve optima.The CPSO-KM utilizes both capability of CPSO K-Means.CPSO-KM has been tested two synthetic datasets three classical data sets from...

Based on a mathematical model involving Radon measure explicitcomputations convolution integrals defining continuous(integral) wavelet transformations are carried out. The studyshows that the truncated Morlet significantly depends arotation parameter and thus lay foundation of edge detection inpattern recognition image processing using rotational(directional) wavelets. Experiments algorithms developedbased theory. theory is further generalized to the$n$-dimensional cases large class rotational

Internet of things devices, namely, radio frequency identification (RFID) tags and sensors deliver the massive lifecycle data agriculture products to database in cloud platform for a supply chain system. The product-related can be efficiently indexed, stored shared among various planting production bases, logistic distribution centers, sales point consumer big environment. This paper proposes an intelligent approach optimize agricultural chains planning by data. Firstly, web crawlers are...

Graph-based semi-supervised node classification has been shown to become a state-of-the-art approach in many applications with high research value and significance. Most existing methods are only based on the original intrinsic or artificially established graph structure which may not accurately reflect "true" correlation among data optimal for downstream neural networks. Besides, while graph-based mostly utilize explicit structure, some implicit information, example, density can also...

Diffusion models have demonstrated strong performance in sampling and editing multi-modal data with high generation quality, yet they suffer from the iterative process which is computationally expensive slow. In addition, most methods are constrained to generate Gaussian noise, limits their flexibility. To overcome both disadvantages, we present Contrastive Optimal Transport Flow (COT Flow), a new method that achieves fast high-quality improved zero-shot flexibility compared previous...

Graph neural networks (GNNs), a class of deep learning models designed for performing information interaction on non-Euclidean graph data, have been successfully applied to node classification tasks in various applications such as citation networks, recommender systems, and natural language processing. is an important research field node-level data mining. Recently, due the limitations shallow GNNs, many researchers focused designing models. Previous GNN architecture search works only solve...

Caused by nonlinear vibration, ocean white noise exhibits complex dynamic characteristics and perception characteristics. To explore the potential application of in engineering health fields, novel methods based on deep learning algorithms are proposed to generate noise, contributing marine environment simulation engineering. A comparative study, including spectrum analysis auditory testing, proved superiority generation method using networks over general mathematical or physical methods....

Motivation: Myocardial Arterial Spin Labeling (myoASL) presents a promising contrast-agent-free approach for assessing myocardial blood flow (MBF), but its clinical translation is hampered by high levels of physiological noise (PN). Goal(s): We introduce double inversion recovery (DIR) preparations FAIR-myoASL to mitigate sensitivity heart rate variations and reduce PN. Approach: A flip-back pulse was added immediately after the FAIR-labeling, allow near-complete and, thus, more effective...