Knowledge Graph (KG) is an essential research direction that involves storing and managing knowledge data, but its incompleteness and sparsity hinder its development in various applications. Knowledge Graph Reasoning ...
Knowledge Graph (KG) is an essential research direction that involves storing and managing knowledge data, but its incompleteness and sparsity hinder its development in various applications. Knowledge Graph Reasoning (KGR) is an effective method to solve this limitation via reasoning missing knowledge based on existing knowledge. The graph Convolution Network (GCN) based method is one of the state-of-the-art approaches to this work. However, there are still some problems, such as the insufficient ability to perceive graph structure and the poor effect of learning data features which may limit the reasoning accuracy. This paper proposes a KGR architecture based on a graph sequence generator and multi-head self-attention mechanism, named GaM-KGR, to improve the above problems and enhance prediction accuracy. Specifically, the GaM-KGR first introduces the graph generation model into the field of KGR for graph representation learning to obtain the hidden features in the data so that enhancing the reasoning effect and then embeds the generated graph sequence into the multi-head self-attention mechanism for subsequent processing to improve the graph structure perception ability of the proposed architecture, so that it can process the graph structure data more appropriately. Extensive experimental results show that the GaM-KGR architecture can achieve the state-of-the-art prediction results of current GCN-based models.
As medical insurance continues to grow in size, the losses caused by medical insurance fraud cannot be underestimated. Current data mining and predictive techniques have been applied to analyze and explore the health ...
详细信息
Machine Learning (ML) has been widely applied to medical science for decades. As common knowledge, the progress of many diseases is often chronic and dynamic. Longitudinal data, or time-series data, has better descrip...
详细信息
Fuzzing is increasingly being utilized as a method to test the reliability of Deep Learning (DL) systems. In order to detect more errors in DL systems, exploring the internal logic of more DNNs has become the main obj...
详细信息
ISBN:
(数字)9798350349184
ISBN:
(纸本)9798350349191
Fuzzing is increasingly being utilized as a method to test the reliability of Deep Learning (DL) systems. In order to detect more errors in DL systems, exploring the internal logic of more DNNs has become the main objective of fuzzing. Despite advancements in the seed selection aspect of fuzzing, considerable opportunities still exist for improving testing efficiency. Current research has issues with the repeated consideration of neurons in the model that will be covered in the future by other seeds, leading to redundant seeds and lower testing efficiency. Additionally, there is a lack of a method to measure the potential of seeds to increase coverage, making it difficult to select the most worthy seeds for mutation in each iteration. We propose an uncovered neurons information based (UNIB) fuzzing method for DNN. UNIB uses clustering methods to organize the seed queue based on initial seed data, aiming to enhance the coverage rate improved in each iteration. It also integrates coverage information from the testing phase to identify the seeds with the greatest potential. The experimental results show that UNIB achieved a higher NC than the second-best method by 1.1% and 3% in LetNet-4 and LetNet-5, respectively. UNIB consistently required the fewest number of iterations to reach the same NC as other methods. For both LetNet-4 and LetNet-5, the adversarial test case sets generated by UNIB exhibited the highest diversity.
Textual network embedding aims to learn meaningful low-dimensional representations for vertices with the consideration of the associated texts. When learning the representations for texts in network embedding, existin...
详细信息
Cryptocurrency phishing scams is a significant treat to Ethereum, one of the most popular blockchain platforms. Most of existing Ethereum phishing detection methods are based on traditional machine learning or graph r...
Cryptocurrency phishing scams is a significant treat to Ethereum, one of the most popular blockchain platforms. Most of existing Ethereum phishing detection methods are based on traditional machine learning or graph representation learning, which mostly rely on only statistical and structural features in local scope. In this paper, we propose Multi-transaction-view Graph Attention Network (MTvGAT), a novel phishing scam detection model that can make use of transaction patterns of different scopes. To obtain global-view information, we apply graph clustering and construct the global-view graph with multiple clusters, including all the nodes of the original transaction network. To obtain local view information, we apply neighborhood sampling, and construct local-view graphs with target nodes and their neighborhood nodes. Then, node features, edge features, and attention coefficients are aggregated to merge multi-view information into representation of nodes. We further combine global-view and local-view representations to finally identify phishing addresses from target nodes. Extensive experiments demonstrate that the proposed method can outperform existing ones with significant improvement.
DNA methylation is of great significance to the diagnosis, treatment and disease prediction of hepatocellular carcinoma (HCC). The commonly used DNA methylation microarrays have high data dimensions, and different CpG...
详细信息
ISBN:
(纸本)9781450384155
DNA methylation is of great significance to the diagnosis, treatment and disease prediction of hepatocellular carcinoma (HCC). The commonly used DNA methylation microarrays have high data dimensions, and different CpG sites detected may map to the same gene. To extract more effective features for HCC disease prediction, this study uses a linear regression model that integrates TCGA database methylation data and gene expression data, based on the DNA methylation microarray data of the GEO database (GSE113017), the corresponding gene expression data was predicted and the differentially expressed genes (3766) with significant differences were screened out, which was used as the feature of the data set. Constructing an Artificial Neural Network (ANN) to train a machine learning model for HCC disease prediction and perform 10-fold cross-validation. The resulting model has an accuracy of 95.1% for HCC disease prediction based on DNA methylation microarray data. Compared with other HCC prediction methods, this method has better performance. Then analyze the differentially expressed genes with protein-protein interaction network (PPI network), and use the top five connected genes in the network as hub genes, namely: GNGT2, GNB4, FPR2, CDC20, NMUR1, which can be used as biomarkers for the diagnosis, treatment and prognosis of HCC.
Kinship verification is an important and challenging task. The aim of it is to find out whether there is a kin relationship between one pair of facial images. However, there are many challenges for kinship verificatio...
详细信息
ISBN:
(纸本)9781450385893
Kinship verification is an important and challenging task. The aim of it is to find out whether there is a kin relationship between one pair of facial images. However, there are many challenges for kinship verification. One of the most important challenge is the problem of data imbalance. Specifically, if there exists N positive kinship pairs, we can get N(N-1) negative pairs. Obviously, the number of negative kinship pairs is much larger than the number of positive kinship pairs. How to make full use of positive and negative samples for training is a valuable problem. But most of the existing methods just randomly select the same number of negative sample pairs as the positive sample pairs. What is more, different negative pairs contain different discrimination information. Therefore, finding discriminative sample is also an important problem. In this paper, we propose a simple and effective method, named Negative sample Mining based Deep Feature Learning (NMDFL) to solve the above problems. Specifically, different with most existing methods, we sample N positive kinship pairs and T*N negative pairs, where T>1. Then, they are sent to the train net, and their impacts on network updating are adjusted in proportion to the number of positive and negative samples. At the same time, the corresponding weight of each negative sample is dynamically adjusted according to the training results of each time. Experimental results on the KinFaceW-I and KinFaceW-II datasets proves the effectiveness of our method.
Few-shot hyperspectral image classification aims to identify the classes of each pixel in the images by only marking few of these pixels. And in order to obtain the spatial-spectral joint features of each pixel, the f...
详细信息
Video surveillance has been using in daily life widely and has significant impact in security field. Reading time information in video has becoming an essential part. Because of the complex background of timestamp and...
详细信息
暂无评论