The detection of abnormal targets in transmission lines plays a significant role in maintaining the stability and safety of transmission systems. To achieve improved detection performance for abnormal targets, we propose a new target detector based on YOLOX, called YOLOX++. First, a multiscale cross-stage partial network (MS-CSPNet) is designed, which fuses multiscale feature information and expands the receptive field of the target through channel combination. Furthermore, depthwise and dilated convolutions are introduced in an object decoupling head to better capture the long-range dependencies of objects in feature maps. Finally, the alpha loss function (α-IoU) is introduced to optimize the localization of small objects. Experiments show that in a comparison with the YOLOX model, the YOLOX++ approach in this paper achieves 86.8% and 96.6% detection accuracies for high-voltage tower bird nest and power line insulator targets, respectively. On the PASCAL VOC dataset, the AP ₅₀ and APS are improved by 9.3% and 5.0% over those of YOLOX, respectively, showing that the YOLOX++ network possesses better robustness for small target detection.

Knowledge base completion (KBC) aims to predict missing information in a knowledge base. Most existing embedding-based KBC models assume that all test entities are available at training time. Thus, a question arises-that is, how to answer queries concerning test entities not observed at training time, which is called the out-of-knowledge-base (OOKB) entity problem. In this paper, we propose a parameter-efficient embedding model that combines the benefits of a graph neural network (GNN) and a convolutional neural network (CNN) to solve the KBC task with OOKB entities. First, we apply the GNN architecture to learn the information between nodes in the graph. Second, convolution layers are used as a transition matrix in GNN to learn more expressive embeddings with fewer parameters. Finally, we use a transition-based knowledge graph embedding model to solve the KBC task. The model has learnable weights that adapt based on information from neighbors and can exploit auxiliary knowledge for OOKB entities to compute their embedding while remaining parameter efficient. We demonstrate the effectiveness of the proposed model on OOKB datasets, and the code is available at https://github.com/Tianchen627/Knowledge-Transfer-for-Out-of-Knowledge-Base-Entities.

With the continuous accumulation of social network data, social recommendation has become a widely used recommendation method. Based on the theory of social relationship propagation, mining user relationships in social networks can alleviate the problems of data sparsity and the cold start of recommendation systems. Therefore, integrating social information into recommendation systems is of profound importance. We present an efficient network model for social recommendation. The model is based on the graph neural network. It unifies the attention mechanism and bidirectional LSTM into the same framework and uses a multilayer perceptron. In addition, an embedded propagation method is added to learn the neighbor influences of different depths and extract useful neighbor information for social relationship modeling. We use this method to solve the problem that the current research methods of social recommendation only extract the superficial level of social networks but ignore the importance of the relationship strength of the users at different levels in the recommendation. This model integrates social relationships into user and project interactions, not only capturing the weight of the relationship between different users but also considering the influence of neighbors at different levels on user preferences. Experiments on two public datasets demonstrate that the proposed model is superior to other benchmark methods with respect to mean absolute error and root mean square error and can effectively improve the quality of recommendations.

With the continuous improvement of the e-commerce ecosystem and the rapid growth of e-commerce data, in the context of the e-commerce ecosystem, consumers ask hundreds of millions of questions every day. In order to improve the timeliness of customer service responses, many systems have begun to use customer service robots to respond to consumer questions, but the current customer service robots tend to respond to specific questions. For many questions that lack background knowledge, they can generate only responses that are biased towards generality and repetitiveness. To better promote the understanding of dialogue and generate more meaningful responses, this paper introduces knowledge information into the research of question answering system by using a knowledge graph. The unique structured knowledge base of the knowledge graph is convenient for knowledge query, can acquire knowledge faster, and improves the background information needed for answering questions. To avoid the lack of information in the dialogue process, this paper proposes the Multi-hop Knowledge Information Enhanced Dialogue-Graph Attention (MKIED-GA) model. The model first retrieves the problem subgraph directly related to the input information from the entire knowledge base and then uses the graph neural network as the knowledge inference module on the subgraph to encode the subgraph. The graph attention mechanism is used to determine the one-hop and two-hop entities that are more relevant to the problem to achieve the aggregation of highly relevant neighbor information. This further enriches the semantic information to provide a better understanding of the meaning of the input question and generate appropriate response information. In the process of generating a response, a multi-attention flow mechanism is used to focus on different information to promote the generation of better responses. Experiments have proved that the model presented in this article can generate more meaningful responses than other models.

Proving termination of program loops is necessary in many applications, especially those safety critical software. Discovering ranking function is a classical method to prove the termination, but the existence of ranking function is a sufficient condition on the termination of a loop program. In this paper, we present an algorithm to remedy the limitation of ranking function for analyzing the termination of linear programs with conditionals. We transform the linear loop programs with conditionals into the nested linear loop programs, and then check whether the inner loop terminates or not by positive eigenvalues and their corresponding eigenvectors. If one of the inner loop in the nested linear loop is nonterminating, then the linear loop is nonterminating. Otherwise, we use discovering ranking functions based on qualifier elimination to analyze the termination of the linear loop. To illustrate the main idea of the algorithm, we perform it on some examples.

Neural network methods have been trained to satisfactorily learn user/product representations from textual reviews. A representation can be considered as a multiaspect attention weight vector. However, in several existing methods, it is assumed that the user representation remains unchanged even when the user interacts with products having diverse characteristics, which leads to inaccurate recommendations. To overcome this limitation, this paper proposes a novel model to capture the varying attention of a user for different products by using a multilayer attention framework. First, two individual hierarchical attention networks are used to encode the users and products to learn the user preferences and product characteristics from review texts. Then, we design an attention network to reflect the adaptive change in the user preferences for each aspect of the targeted product in terms of the rating and review. The results of experiments performed on three public datasets demonstrate that the proposed model notably outperforms the other state-of-the-art baselines, thereby validating the effectiveness of the proposed approach.