A single model usually cannot learn all the appropriate features with limited data, thus leading to poor performance when test data are used. To improve performance, we propose a teacher-student collaborative knowledge distillation (TSKD) method based on and self-distillation. The consists of two parts: learning in teacher network self-teaching student network. Learning allows use from Self-teaching is build multi-exit self-distillation provide deep as supervised information for training. In...

Abstract This study examines whether executives justify their excess compensation through environmental information disclosure using a sample of listed companies in China's heavily polluting industries from 2010 to 2014. We find that executives' is positively related the quality disclosure. The above relationship significant cases with strong demand for justification (i.e., state‐owned enterprises and firms where an internal gap salient), indicating manipulate about according purposes. also...

The extant research on supply chain information sharing under wholesale pricing often assumes a model where the manufacturer can unilaterally set any price, and retailer decides retail quantity or price while taking as given. Whereas this may actually reflect relative market power in some situations, its implementation misses out certain win–win opportunities that could arise from sharing. We propose new mechanism promotes between relatively weaker more powerful such both parties become...

Entity alignment (EA), a crucial task in knowledge graph (KG) research, aims to identify equivalent entities across different KGs support downstream tasks like KG integration, text-to-SQL, and question-answering systems. Given rich semantic information within KGs, pre-trained language models (PLMs) have shown promise EA due their exceptional context-aware encoding capabilities. However, the current solutions based on PLMs encounter obstacles such as need for extensive training, expensive...

ABSTRACT Although many studies have conducted the traffic scheduling of time‐sensitive networks, most focus on small‐scale static for specific scenarios, which cannot cope with dynamic and rapid time‐triggered (TT) flows generated in scalable scenarios Industrial Internet Things. In this paper, we propose a Scalable TT flow method based Dynamic Online Grouping industrial networks (SDOG). To achieve that, establish an undirected weighted graph conflict index between divide available time into...

Tuning hyperparameters is a crucial but arduous part of the machine learning pipeline. Hyperparameter optimization even more challenging in federated learning, where models are learned over distributed network heterogeneous devices; here, need to keep data on device and perform local training makes it difficult efficiently train evaluate configurations. In this work, we investigate problem hyperparameter tuning. We first identify key challenges show how standard approaches may be adapted...

Abstract Purpose . Real-time three-dimensional (3D) magnetic resonance (MR) imaging is challenging because of slow MR signal acquisition, leading to highly under-sampled k-space data. Here, we proposed a deep learning-based, k-space-driven deformable registration network (KS-RegNet) for real-time 3D imaging. By incorporating prior information, KS-RegNet performs image between fully-sampled and on-board images acquired from highly-under-sampled data, generate high-quality motion tracking....

Resource-based companies are key players in reducing carbon emissions and play a central role achieving China’s dual-carbon goal. Establishing improving an objective information disclosure mechanism for evaluating the quality of scientific reasonable manner have significant reference value rationally shaping way to realize peak neutrality. In view this, this paper develops evaluation index system based on four dimensions corporate social responsibility reports listed from 2018 2022. After...

Not Available For Download Add Paper to My Library Share: Permalink Using these links will ensure access this page indefinitely Copy URL DOI

The extant research on supply chain information sharing under wholesale pricing often assumes a model where the manufacturer can unilaterally set any price, and retailer decides retail quantity or price while taking as given. Whereas this may actually reflect relative market power in some situations, it misses out certain win-win opportunities that could arise from sharing. When demand shared by is used to more advantageous for its own benefit only, double marginalization effect will be...

Abstract This paper researches the pricing strategy and government intervention mechanism for green supply chain in monopoly market, while considering strategic customer behavior. According to optimization theory, it establishes target functions of retailer customers investigates interactions between accordance with Stackelberg’s game so as confirm optimal discount level products. In addition, discusses regulatory effect intervention, including guiding price fiscal subsides, on sales The...

In-service inspection is one of the important means to ensure nuclear safety, and current manual management model has been difficult adapt high requirements power development. For example, there have problems such as inconsistency key data information, difficulty in statistical analysis results, lack or untimely experience feedback. Through research application in-service technology platform, it focuses on platform design, business architecture, technologies actual development process. The a...

The development of modern technology and e-commerce have given rise to the emergence many new selling channels. Among one them, group-buying attracts numerous customers rapidly due characters deep discounts great convenience. Although create sales growth for sellers, it also causes loss in their profit margins. Meanwhile, business model websites is not thoroughly understood literature. Based on a Stackelberg game framework, this paper studies equilibrium between website seller. optimal...

Image segmentation has always been the difficulty and focus in image processing research. The traditional best histogram threshold method is not ideal for of images. This paper presents a modified two-dimensional entropy its improved genetic algorithm. Compared with method, only reflects gray-level distribution information, but also pixel neighborhood spatial correlation greatly improving effect. At same time, algorithm used to solve problem that slow convergence improper selection. improves...

With the development of agricultural modernization, traditional production model has been gradually subverted, and planting is no longer main way to increase farmers' income. Under mode, farmland management control mainly rely on experience, fertilization technology, etc., which cannot adapt modern environment, improve efficiency, promote Based application mobile intelligent platforms, real-time information collection, storage analysis can be realized. In order reduce use chemical fertilizer...

Neural Radiance Field (NeRF) has received much attention in recent years due to the impressively high quality 3D scene reconstruction and novel view synthesis. However, image degradation caused by scattering of atmospheric light object particles atmosphere can significantly decrease when shooting scenes hazy conditions. To address this issue, we propose Dehazing-NeRF, a method that recover clear NeRF from inputs. Our simulates physical imaging process images using an model, jointly learns...