Search

Data trading platforms play a crucial role in facilitating data circulation and promoting the sustainable allocation of data resources. Establishing a transparent, fair, and efficient pricing mechanism is key to ensuring the long-term stability and development of such platforms. However, these platforms face challenges in pricing due to the small sample problem, as traditional machine learning methods typically rely on large amounts of data. To address this issue, this paper proposes a data resource pricing model that combines WGAN-GP data augmentation and the Reptile algorithm. Data augmentation generates related datasets to increase sample size, enhancing the renewability of data resources, while meta-learning transfers knowledge across tasks, improving the model's ability to quickly adapt to new tasks and efficiently utilize resources. Validation using actual trading data from the data trading platform shows that the proposed model accurately predicts data resource prices under small-sample conditions, outperforming other models. This study addresses the limitations of existing pricing methods in small-sample scenarios, providing a sustainable pricing solution for small-sample data resources and improving the accuracy and long-term stability of data pricing in the market. [ABSTRACT FROM AUTHOR]

In order to achieve the recognition of grain direction of wood beam column timber, a few-shot wood grain direction detection method based on the improved meta-learning Reptile algorithm is proposed for the problems of high requirements of image acquisition environment and poor robustness of traditional image processing methods. The model calibration part of the traditional Reptile algorithm is improved by removing the last classification layer of the neural network, and the data from the support set in the training set are randomly divided into two groups, and the network is further fine-tuned using the feature expressions of these two groups and the self-learning cosine similarity measure so that the input data can be more fully learned. A strong theoretical support is provided for the study of wood grain orientation recognition. Experimental results show that the improved Reptile algorithm can recognize the direction of wood texture with an accuracy of 86.6 % under 10-shot. In addition, the accuracy of the proposed algorithm has all improved compared with the traditional deep learning, transfer learning and the Reptile algorithm before improvement in the cases of 1-shot, 5-shot, 10-shot and 100-shot. [ABSTRACT FROM AUTHOR]

In a distribution system, sparse reliable samples and inconsistent fault characteristics always appear in the dataset of neural network fault detection models because of high impedance fault (HIF) and system structural changes. In this paper, we present an algorithm called Generative Adversarial Networks (GAN) based on the Reptile Algorithm (GANRA) for generating fault data and propose an evolution strategy based on GANRA to assist the fault detection of neural networks. First, the GANRA generates enough high-quality analogous fault data to solve a shortage of realistic fault data for the fault detection model’s training. Second, an evolution strategy is proposed to help the GANRA improve the fault detection neural network’s accuracy and generalization by searching for GAN’s initial parameters. Finally, Convolutional Neural Network (CNN) is considered as the identification fault model in simulation experiments to verify the validity of the evolution strategy and the GANRA under the HIF environment. The results show that the GANRA can optimize the initial parameters of GAN and effectively reduce the calculation time, the sample size, and the number of learning iterations needed for dataset generation in the new grid structures.

With the rapid development of virtual communities, more and more customers participate in product innovation and knowledge sharing through virtual communities. Research on virtual community members, especially for community members, will help the community to manage the members and further promote community development and knowledge innovation. At present, the main difficulty in the study of community member user portraits lies in the user's grasp of user behavior data in the community. There is a large amount of structured data and semi-structured data in the community, which is crucial for the portrayal dimension of user portraits. This paper uses the association rule crawler algorithm to conduct community user behavior data association search, and uses text mining, social network analysis (SNA) and clustering technology to image users in the knowledge innovation community from the perspectives of professionalism, participation enthusiasm and network capability. The main result divides users into fancier, participator, and tourist.