GraphCut algorithm has been widely utilized to solve various types of computer vision problems. Its expensive computational cost encouraged many researchers to improve the speed of the algorithm. Recent works proposed...
详细信息
ISBN:
(纸本)9781665452199
GraphCut algorithm has been widely utilized to solve various types of computer vision problems. Its expensive computational cost encouraged many researchers to improve the speed of the algorithm. Recent works proposed schemes that work on parallel computing platforms such as CUDA. However, the problem of low convergence speed prevents the usage of GraphCut for real time applications. In this paper, we propose global suppression heuristic to boost the convergence process of the algorithm. A parallel implementation of GraphCut algorithm on CUDA designed for the image stitching problem is introduced. Our method achieves up to 3× time boost on the graph of size 80×480 compared to the best sequential GraphCut algorithm while achieving satisfactory stitched images, suitable for panorama applications. Our source code will be soon available for further research.
Automatic modulation classification (AMC) is a key step of signal demodulation that determines whether the receiver can correctly receive the modulation type of the transmitted signal without prior knowledge. Deep lea...
详细信息
ISBN:
(纸本)9781665452199
Automatic modulation classification (AMC) is a key step of signal demodulation that determines whether the receiver can correctly receive the modulation type of the transmitted signal without prior knowledge. Deep learning (DL) based AMC methods have achieved excellent performances. However, these methods highly rely on expert experience to design network structures. These hand-designed networks have fixed structures and lack flexibility, which often leads to insufficient model generalization. Neural architecture search (NAS) is a vital direction for automatic machine learning (AutoML) which can solve the shortcomings of hand-designed networks. In this paper, we propose a lightweight progressive differentiable architecture search-based AMC (PDARTS-AMC) method to search for a very lightweight network with good performance. Experimental results show that the proposed PDARTS-AMC method both improves the accuracy and reduces the computational cost when compared with existing methods.
Internet technology has facilitated the development of social media, which has become the main way for people to get news. The low cost of information generation has allowed unconfirmed fake news to be noticed and rep...
详细信息
ISBN:
(纸本)9781665452199
Internet technology has facilitated the development of social media, which has become the main way for people to get news. The low cost of information generation has allowed unconfirmed fake news to be noticed and reposted, which can easily cause social and economic harm. Much of the existing research on fake news detection has focused solely on the accuracy of detection, but has neglected the time efficiency of detection. We propose a fake news detection method called MFEAST, which simplifies article content and speeds up model training by extracting key information from articles. Compared with existing real fake news detection methods, our method substantially reduces the training time and achieves high detection accuracy.
As an economic commodity, data sharing, circulation and trading can not only reduce the maintenance and management costs of enterprises, but also tap the potential value of data, improve the internal workflow of enter...
详细信息
ISBN:
(纸本)9781665452199
As an economic commodity, data sharing, circulation and trading can not only reduce the maintenance and management costs of enterprises, but also tap the potential value of data, improve the internal workflow of enterprises and the cooperation among enterprises. The marketization of data elements and the clarification of data sovereignty are the current difficulties hindering data flow. This paper addresses one of the current data circulation issues: how to maintain data sovereignty, and makes exploration and research in combination with the current era background. For the current research projects and products, compare and analyze the techniques used to maintain data sovereignty. Finally, based on the current technology, it gives recommendations for the future development of data sovereignty protection technology.
The construction of heterogeneous social networks enables the major social platforms in the network to connect through social information. In order to ensure network security and improve downstream tasks such as user ...
详细信息
ISBN:
(纸本)9781665452199
The construction of heterogeneous social networks enables the major social platforms in the network to connect through social information. In order to ensure network security and improve downstream tasks such as user profile, knowledge graph construction and recommendation, the relevance measurement between social information has attracted extensive attention in recent years. Although HeteSim algorithm has achieved good results in measuring the relevance between heterogeneous nodes, this method only focuses on the structure features between nodes, and fails to comprehensively consider the joint impact of structure features and semantic features. Therefore, this paper proposes HeteSim-Measured algorithm that considers the fusion of structure features and semantic features for improving the accuracy of relevance measurement. The experiment is verified by measuring the relevance based on meta-path on the datasets and comparing with HeteSim algorithm.
This paper utilizes the finite-time passivity (FTP) to settle the finite-time consensus (FTC) problem for a class of nonlinear fractional-order multi-agent systems (FOMASs). By selecting appropriate state feedback con...
详细信息
ISBN:
(纸本)9781665452199
This paper utilizes the finite-time passivity (FTP) to settle the finite-time consensus (FTC) problem for a class of nonlinear fractional-order multi-agent systems (FOMASs). By selecting appropriate state feedback controller, a FTP criterion for FOMAS is derived on the basis of inequality techniques, and a sufficient condition to ensure FOMAS can achieve FTC is also given by exploiting the FTP result. Moreover, to testify the validity of the FTP and FTC criteria, a numerical example is presented.
With the increasing awareness of privacy protection and data security, people’s concerns over the confidentiality of sensitive data still limit the application of distributed artificial intelligence. In fact, a new e...
详细信息
ISBN:
(纸本)9781665452199
With the increasing awareness of privacy protection and data security, people’s concerns over the confidentiality of sensitive data still limit the application of distributed artificial intelligence. In fact, a new encryption form, called homomorphic encryption(HE), has achieved a balance between security and operability. In particular, one of the HE schemes named Paillier has been adopted to protect data privacy in distributed artificial intelligence. However, the massive computation of modular multiplication in Paillier greatly affects the speed of encryption and decryption. In this paper, we propose a fast CRT-Paillier scheme to accelerate its decryption process. We first introduce the Montgomery algorithm to the CRT-Paillier to improve the process of the modular exponentiation, and then compute the modular exponentiation in parallel by using OpenMP. The experimental results show that our proposed scheme has greatly heightened its decryption speed while preserving the same security level. Especially, when the key length is 4096-bit, its speed of decryption is about 148 times faster than CRT-Paillier.
Multimodal aspect-level sentiment classification aims to utilize images to recognize the sentiment polarity of target aspects in text. To address the issues of low utilization of inter-modal complementary information ...
详细信息
ISBN:
(纸本)9781665452199
Multimodal aspect-level sentiment classification aims to utilize images to recognize the sentiment polarity of target aspects in text. To address the issues of low utilization of inter-modal complementary information and vanishing gradients, a multimodal aspect-level sentiment based on multi-selection attention mechanism is proposed. Multi-selection attention mechanism explicitly considers the contribution of different modalities to aspects and utilizes shared features and private features of image modality to enhance sentiment expression of target aspects. On this basis, inspired by residual connections in ResNet and encoder-decoders in U-Net, a simple and effective residual encoder-decoder is proposed to mine deep information and avoid vanishing gradients. The experimental results on two public sentiment datasets show that the proposed model can better utilize images to supplement textual modality and improve the accuracy of sentiment classification.
Graph Neural Networks (GNNs) are effective models for processing graph-structured data. With the continuous growth of graph data scale and the deepening of graph neural network layers, the heavy cost of GNN inference ...
详细信息
ISBN:
(纸本)9781665452199
Graph Neural Networks (GNNs) are effective models for processing graph-structured data. With the continuous growth of graph data scale and the deepening of graph neural network layers, the heavy cost of GNN inference has greatly limited its application in real-time tasks. This paper focus on accelerating the performance of GNN inference. We first measures the execution time ratio of each stage in the inference process for commonly used GNN models, and analyzes the performance characteristics of different stages. We find out that the critical performance factor of GNN inference is the feature dimension, which is different to CNN and NLP models. Therefore, we propose a soft channel pruning method with a ladder pruning pattern. It reduces the calculation from unimportant graph node features and achieve performance acceleration. Meanwhile, it reserves inference accuracy of GNNs. According to experimental validation on graph datasets of different scales, our method can effectively reduce the inference latency and achieve 2×–7× speedup. Also, compared with existing pruning methods, higher inference accuracy can be obtained with comparable speedup ratio.
In multi-view partial multi-label learning (MVPML) problem, each instance is described by several heterogeneous feature representations and associated with a set of candidate labels, which include both ground-truth an...
详细信息
ISBN:
(纸本)9781665452199
In multi-view partial multi-label learning (MVPML) problem, each instance is described by several heterogeneous feature representations and associated with a set of candidate labels, which include both ground-truth and noisy labels. The key to learn from MVPML data lies in how to deal with multi-view data and how to select the ground-truth labels from candidate label set. In this paper, we propose a Graph-based Multi-view Partial Multi-label method, which integrates exploiting multi-view information, noisy label disambiguation and training predictor model into a whole framework. Specifically, we first exploit the consensus information across different views by learning the similarity graph of each view and fuses these similarity graphs into a unified graph. Secondly, we decompose the observed label set into a ground-truth label matrix and a noisy label matrix, where the noisy label matrix is assumed to be sparse. Then, we embed the learned unified similarity graph into the process of label disambiguation to obtain a more reliable ground-truth label matrix. Finally, the predictive model is learned by the ground-truth label matrix. Extensive experiments indicate that our proposed method can achieve superior or comparable performance against state-of-the-art methods.
暂无评论