Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2022-05-18 Cooperative journals: 《计算机应用研究》
Abstract： Video super-resolution (VSR) , which aims to exploit information from multiple adjacent frames to generate a high-resolution version of a reference frame. Many existing VSR works focus on how to effectively align adjacent frames to better fuse adjacent frame information, and little research has been done on the important step of adjacent frame information fusion. To solve this problem, This paper propose a video super-resolution model based on group feedback fusion mechanism (GFFMVSR) . Specifically, after adjacent frames are aligned, the aligned video sequences are fed into the first temporal attention module. Then, the sequence is divided into several groups, and each group achieves preliminary fusion through the intra-group fusion module in turn. Next, the fusion results of different groups go through a second temporal attention module. Then, each group inputs the feedback fusion module group by group, and uses the feedback mechanism to feedback and fuse the information of different groups. Finally, the fusion result output is reconstructed. It has been verified that the model has strong information fusion ability, and is superior to the existing models in both objective evaluation indicators and subjective visual effects.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2022-05-10 Cooperative journals: 《计算机应用研究》
Abstract： In order to improve the accuracy of text classification and solve the problem of insufficient utilization of node features by text graph convolution neural network, this paper proposes a new text classification model, which integrates the advantages of text graph convolution and Stacking integrated learning method. The model first learns the global expression of documents and words and the grammatical structure information of documents through text graph convolution neural network, and then secondary learns the features extracted by text graph convolution through integrated learning, so as to make up for the insufficient utilization of text graph convolution node features, and improve the accuracy of single label text classification and the generalization ability of the whole model. In order to reduce the time consumption of ensemble learning, the fusion algorithm removes the k-fold cross verification mechanism in ensemble learning. The fusion algorithm realizes the correlation between text graph convolution and stacking integrated learning method. The classification effect on R8, R52, Mr, Ohsumed, 20ng and other data sets is improved by more than 1.5%, 2.5%, 11%, 12% and 7% respectively compared with the traditional classification model. This method performs well in the comparison of classification algorithms in the same field.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2022-04-07 Cooperative journals: 《计算机应用研究》
Abstract： Recommendation systems are widely used everywhere and have a great influence on daily life. Aiming to train an ideal recommendation system, a massive of ‘use-item’ interactive pairs should be provided. However, the dataset obtained is usually sparse, which might result in an over-fitting model and be hard to obtain the satisfying performance. In order to address this problem, the cross-domain recommendation is raised. Most of the existing methods on cross-domain recommendation systems borrow the ideas of the conventional unsupervised domain adaptation, which employ the feature alignment or adversarial training methods. Hence they can transfer the domain-invariant interests of users from the source to the target domains, e. g. , from the Douban Movies to the Douban Books. However, since the network structures vary with different recommendation platforms, the existing methods on cross-domain recommendation systems straightforwardly extract the domain-invariant representation may entangle the structure information, which may result in the false alignment phenomenon. Furthermore, the previous efforts ignore the noise information behind the graph data, which further degenerate the experimental performance. In order to address the aforementioned problems, this paper brings the causal generation process of graph data into the cross-domain recommendation systems, this paper use the semantic latent variables of each node to calculate the relationships between users and items via disentangling the semantic latent variables, domain latent variables and noise latent variables. Experiments studies testify that the proposed method yields state-of-the-art performance on several cross-domain recommendation system benchmark datasets.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2020-09-28 Cooperative journals: 《计算机应用研究》
Abstract： This paper proposed a multiscale feature fusion approach for malicious HTTP request detection. Firstly, it models the HTTP request in both word-level and character-level. Secondly, it extracts the high level sematic information in HTTP request by using a specially designed convolutional neural network (CNN) . Thirdly, it jointly learns the multiscale representation for HTTP request with the help of multimodal learning techniques. Finally, a linear classifier is adopted for classification. Extensive experiments conducted on public HTTP CSIC 2010 dataset and WAF dataset show large improvement on the performance against existing state-of-the-art methods.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2020-09-28 Cooperative journals: 《计算机应用研究》
Abstract： This paper studied a new event-triggered consensus protocol for MASs based on switching topology. Two events are designed in this protocol, including convergence events caused by agent state evolution and topology events caused by topology change. Stability analysis of public Lyapunov function related to events and consensus proof of MASs are given. A network connectivity algorithm based on constraint set is also designed to enhance network connectivity. The simulation results show that the designed event-triggered consensus protocol can make the state evolution of agents tend to be consistent and effectively reduce the update frequency of the controller of agents in the switching topology network environment, reduce the energy consumption of MASs, improve the connectivity effect of the network effectively, and provide theoretical support for the subsequent research of event-triggered control.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2019-05-10 Cooperative journals: 《计算机应用研究》
Abstract： For text matching problems in natural language processing, this paper proposed a deep learning model based on self-adaptive affinity graph learning framework for short text matching. The affinity graph can be converted into a vector form using word embedding, and then obtained by constructing a text similarity relationship matrix, which can express the neighbor relationship of the text sample. Current methods usually construct static affinity graphs, which rely on prior knowledge and hard to obtain the optimal representation of sentence pairs. Therefore, this paper proposed to use the Siamese CNN to learn the affinity graph of better dynamic updates. The accuracy and F1 values of the model on the Quora dataset are 84.15% and 79.88%, respectively, and the accuracy and F1 values on the MSRP dataset are 74.55% and 81.63%, respectively. Experiments show that the proposed model can improve the accuracy of text recognition and matching effectively.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2019-04-01 Cooperative journals: 《计算机应用研究》
Abstract： So far, many different machine learning methods have been proposed, and traditional machine learning methods can not effectively solve the classification problem of large-scale intrusion data. In order to solve the problem of classification of large-scale intrusion data, This paper proposed lightGBM binary tree algorithm based on stacked sparse autoencoder. Firstly, the category labels were divided into five categories and constructed into binary tree structures, then the imbalance of data distribution was solved by the upper sampling method, the above processing could separate the large-scale data, so that they could be trained separately, and then, the sparse autoencoder network was used to reduce the feature dimension. Using this method could ensure that time of dimension reduction could be saved on the basis of extracting deeper features from the original data. Finally, the lightGBM ensemble algorithm was used to classify. And compared to other models, using the lightGBM model could save training time while ensuring classification performance. The NSL-KDD dataset was used to measure the accuracy, accuracy, recall, and comprehensive evaluation index F1 of the proposed method, which reached an average of 87.42 %, 98.20 %, and 91.31 % in five classification, respectively. It is superior to the comparison algorithm and obviously saves the calculation time.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2019-01-28 Cooperative journals: 《计算机应用研究》
Abstract： Due to the complexity and variety of glyphs, offline handwritten Chinese character recognition has always been a difficult problem of pattern recognition. The development of deep convolutional neural networks provides a direct and effective solution to this problem. This paper studied offline handwritten Chinese character recognition based on Inceptions neural network.It proposed an improved Inception structure, which took the advantages of simpler structure, easier network depth expansion and less training parameters. The method used the proposed structure to verifiy on dataset CISIA-HWDB1.1. The model achieved an average accuracy of 96.95%, by using stochastic gradient descent optimization algorithm. Experimental result shows that the improved Inception structure has better generalization performance and robustness in image classification, and can be easily extended to other applications.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2019-01-28 Cooperative journals: 《计算机应用研究》
Abstract： The existing mainstream methods use the pre-trained convolutional neural networks to extract image features and usually have the following limitations: a）Only using a single layer of pre-trained features to represent image; b）Inconsistency between the pre-trained task and the actual research task. These limitations result in that the existing methods of image-text matching cannot make full use of image features and is easily influenced by the noises. To solve the above limitations, this paper used multi-layer features from a pre-trained network and proposed a fusion algorithm of multi-level image features accordingly. Under the guidance of the image-text matching objective function, the proposed algorithm fused the multi-level pre-trained image features and reduced their dimensionality using a multi-layer perceptron to generate fusion features. It is able to make full use of pre-trained features and successfully reduce the influences of noises. The experimental results show that the proposed fusion algorithm makes better use of pre-trained image features and outperforms the methods using single-level features in the image-text matching task.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2019-01-28 Cooperative journals: 《计算机应用研究》
Abstract： In view of the existing RFID authentication protocol is insufficient in the security authentication process due to the flaws in the design of the protocol, this paper presented an improved lightweight RFID authentication protocol using synchronized random number and PUF. It first proposed a desynchronizing attack method on RFID protocols, and analyzed the reason.Then it enhanced the protocol's ability to resist desynchronization attacks by setting a synchronized random number at both ends of the tag and reader.Finally, it introduced the PUF in the tag, used the PUF’s non-clonality to improve the tag’s anti-attack capability. The analysis results show that the new protocol can effectively resist multiple attacks, and it has higher security while ensuring certain efficiency and overhead.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2019-01-28 Cooperative journals: 《计算机应用研究》
Abstract： The interest flooding attack (IFA) in content-centric networking (CCN) is a hot topic of research on CCN network security. In order to improve the ability of CCN to defend against IFA, this paper studied different defense methods and proposed an improved defense method for interest flooding attack in CCN. According to the CCN flow balance principle, the method used the method of malicious prefix tracing to achieve fast detection of IFA, and the defense against IFA was achieved by improving the Additive Increase Multi-plicative Decrease (AIMD) algorithm. The security analysis shows that the proposed method can defend faster when it in the face of IFA. Compared with other defense methods, this method reduces the computational cost of the CCN router when detecting IFA under the premise of ensuring security.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2019-01-03 Cooperative journals: 《计算机应用研究》
Abstract： At present, the existing convolutional neural network has complicated structure and bases on huge dataset. So it has difficulties in meeting the requirement of computing performance and limitation of energy consumption requested by some practical applications or computing platforms. We studied the binary algorithm based on ARM+ FPGA platform and designed a binary neural network aiming at these applications or platforms. This work reduces the demand for data storage units and simplifies the computational complexity. When implemented in the ARM+ FPGA platform, the convolution multiply-accumulate operation is converted into XNOR logic and popcount operation, which improves the overall operation efficiency and declines the consumption of energy and resources. At the same time, based on the characteristics of data storage in binary neural network, a new row processing algorithm is proposed to improve the throughput of the network. In a word, This implementation is superior to the existing FPGA neural network acceleration methods in terms of GOPS, energy and resource efficiency.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-12-13 Cooperative journals: 《计算机应用研究》
Abstract： Aiming at the high cost of obtaining the training data set, proposing a new weak supervision method for image saliency detection. Only using the picture-level label when training the network model. Dividing the method into two stages. In the first stage, training the classification model according to the picture-level label to obtain the foreground inference graph. In the second stage, processing the original image by super-pixel block and merged with the foreground inference graph obtained in phase one, thus refining significant object boundaries. The algorithm uses existing large training sets and image-level tags, eliminating the use of pixel-level tags, which reduces the amount of annotation work. The experimental results on the four common benchmark datasets show that the performance is significantly better than the unsupervised model, and it has certain advantages compared with the full-supervised model.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-12-13 Cooperative journals: 《计算机应用研究》
Abstract： Concern the problem that the traditional association analysis algorithms cannot efficiently and accurately mine the user's potential temporal association control habits which are implied in the user's operation records, this paper proposed a novel user temporal association control habits mining method based on FP-Growth. This method includes three stages: to generate the transaction set, the temporal frequent item set, and the final temporal association control habits via the user operation-action forest, the improved FP-Growth algorithm and a time constraint rule. Finally, the comparative experiments by using the real user control records show that this method can improve the efficiency of transaction set generation and can more accurately discover the user’s temporal association habits of smart home devices.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-11-29 Cooperative journals: 《计算机应用研究》
Abstract： In order to solve the problems existing in the learning recommendation algorithm that ignore the analysis of the students' knowledge points and can not probabilize the knowledge mastery, this paper proposed a recommendation method based on multiple factors. The method focused on the comprehensive weight of knowledge points, error rate and loss rate, and built a knowledge point mastery probability model, and applied the proposed strategy to implement an online personalized learning recommendation system . In terms of the systematic evaluation, through a survey of 200 high school students, the accuracy of the top-8 knowledge points recommended by our system achieves significant performance, Precision: 91.2%, and F1: 78.4％. The results of the systematic survey reflect the effectiveness and reliability of the proposed strategy.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-11-29 Cooperative journals: 《计算机应用研究》
Abstract： The method based on recurrent neural network combined with syntactic structure is widely used in relation classification, and the neural network is used to automatically acquire features and realize relation classification. However, the existing methods are mainly based on a single specific syntactic structure model, and the model of a specific syntactic structure cannot be transferred to other types of syntactic structures. Aiming at this problem, a hierarchical recurrent neural network model with multi-syntactic structure is proposed. The hierarchical recurrent neural network is divided into two layers for network construction. Firstly, entity pre-training is performed in the sequence layer. The Bi-LSTM-CRF fusion Attention mechanism is used to improve the model's attention to the entity information on the text sequence, thereby obtaining more accurate. The more accurate entity feature information promotes better classification in the relation layer stage. Secondly, in the relation layer, the Bi-Tree-LSTM is nested above the sequence layer, and the hidden state and entity feature information of the sequence layer is passed into the relation layer, then three different syntax structures are weighted learned using the shared parameters and classify the semantic relation finally. The experimental results show that the model has a marco-F1 value of 85.9% on the SemEval-2010 Task8 corpus, and further improves the robustness of the model.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-11-29 Cooperative journals: 《计算机应用研究》
Abstract： Feature selection technology plays an important role in big data analysis, image processing, bioinformatics and other fields. In practical applications, the objectives of reducing the classification error rate and reducing the number of extracted features for facilitating the use of subsequent data, are often two conflicting goals. The multi-object particle swarm optimization based on crowding, mutation, dominance for feature selection (CMDPSOFS) is a kind of bi-objective optimization algorithm with the minimal number of features and classification error rate in feature-oriented selection applications. The algorithm uses three different mutation mechanisms for maintaining swarm diversity and balancing global and local search capabilities. However, the uniform variation increases the randomness of the algorithm, resulting in the generation of worse solutions, which reduces the convergence speed of the algorithm. This paper proposed an improved CMDPSOFS-II algorithm to introduce the mutation and selection operations of differential evolution algorithm into the CMDPSOFS algorithm. The experimental results show that the CMDPSOFS-II algorithm is superior to the original method in feature selection and better balances global and local search capabilities.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-11-29 Cooperative journals: 《计算机应用研究》
Abstract： Convolutional neural network's high computational complexity often obstructs its widespread adhibition in real-time and low-power applications. The existing software implementation solution cannot meet the demands of the convolutional neural network for computing performance and power consumption. The traditional FPGA-oriented convolutional neural network construction method has problems such as complicated process, long cycle and small optimization space. For these problems, according to the characteristics of convolutional neural network calculation pattern, this paper proposed a design and scheduling mechanism of convolutional neural network accelerator for cloud FPGAs. By using for reference the design which based HLS technology, imported the cyclic cutting parameters and rearranged the convolution layer circularly. Then constructed the network in a modular way, and extended parameters to further optimize the accelerator processing process. Summarized the scheduling scheme by analyzing the characteristics of system tasks and resources, and optimized its design from two aspects of control and data flow. In comparison with other existing work, the proposed design provided a solution with flexibility, low energy consumption, high energy efficiency and performance. The design also discussed the efficient universal scheduling scheme of the accelerator. Experimental results show that compared with the CPU implementation, this design achieves 8.84 times speedup of AlexNet, while the power consumption of Cifar implementation is only 24.96% of it. Compared with the CPU+GPU to achieve 6.90 times speedup of Cifar, although the performance of large-scale network is inferior to the GPU, but the minimum power consumption is only 14.98%. This design achieves the maximum acceleration of 6.29 times in comparison with the existing research results. Compared to the accelerators generated for large platforms, even if it only has comparable performance but with a lower clock frequency.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-11-29 Cooperative journals: 《计算机应用研究》
Abstract： In order to improve the clustering accuracy of hyperspectral images, this paper proposed a new sparse subspace clustering model combined three-dimensional spatial spectral features with subspace clustering algorithms. While focusing on the spectral information of hyperspectral images, it also paid attention to spatial context information. Firstly, extracting three kinds of three-dimensional spatial spectral features from the pixels of the hyperspectral image. Then the features weighted the coefficient matrix of the subspace clustering model so that the pixel points could sparsely represent the pixel point to which they are most similar, thereby obtaining the better coefficient matrix. Finally, it used the coefficient matrix to obtain better clustering results with spectral clustering. The algorithm experimented on four classical hyperspectral datasets, and compared the experimental results with six clustering algorithms. The results show that the proposed algorithm achieves higher clustering accuracy on the four datasets than the other algorithms. The algorithm can achieve at most 8.62% accuracy than the algorithms based 3D spatial spectral features like M3DF^3 algotithm and 3DF-SSC algorithm, and at most 25.18% than the algorithms which improve the subspace clustering algorithm by using spatial context information like L2-SSC algorithm and SS-LRSC algorithm.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-10-11 Cooperative journals: 《计算机应用研究》
Abstract： Existing Proof of Ownership schemes are vulnerable to the threat of honest-but-curious servers. With the help of a trusted third party to solve the problem can lead to high overhead. This paper proposed an improved Proof of Ownership scheme based on Dynamic Bloom Filter, which doesn’t require a trusted third party. The scheme uses convergent encryption against honest-but-curious servers. To resist data poisoning attack, it checks whether the encrypted blocks correspond with the tokens at server side. Moreover, a key chaining mechanism is used to solve the problem that convergent keys require too much storage space in existing schemes. Analyses and comparisons show that the scheme has lower key storage overhead and transferring overhead.
Hits
Hits
Downloads
Comment