检索结果-内蒙古大学图书馆

Research and application of Transformer based anomaly detection model: A literature review

学校读者我要写书评

暂无评论

arXiv

arXiv 2024年

作者： Ma, Mingrui Han, Lansheng Zhou, Chunjie Hubei Key Laboratory of Distributed System Security Hubei Engineering Research Center on Big Data Security School of Cyber Science and Engineering Huazhong University of Science and Technology Hubei Wuhan430074 China The Key Laboratory of Ministry of Education for Image Processing and Intelligent Control School of Artificial Intelligence and Automation Huazhong University of Science and Technology Hubei Wuhan430074 China

Transformer, as one of the most advanced neural network models in Natural Language processing (NLP), exhibits diverse applications in the field of anomaly detection. To inspire research on Transformer-based anomaly detection, this review offers a fresh perspective on the concept of anomaly detection. We explore the current challenges of anomaly detection and provide detailed insights into the operating principles of Transformer and its variants in anomaly detection tasks. Additionally, we delineate various application scenarios for Transformer-based anomaly detection models and discuss the datasets and evaluation metrics employed. Furthermore, this review highlights the key challenges in Transformer-based anomaly detection research and conducts a comprehensive analysis of future research trends in this domain. The review includes an extensive compilation of over 100 core references related to Transformer-based anomaly detection. To the best of our knowledge, this is the first comprehensive review that focuses on the research related to Transformer in the context of anomaly detection. We hope that this paper can provide detailed technical information to researchers interested in Transformer-based anomaly detection tasks. Copyright © 2024, The Authors. All rights reserved.

关键词： Anomaly detection

Collaborative Bi-Aggregation for Directed Graph Embedding

学校读者我要写书评

暂无评论

SSRN 2023年

作者： Liu, Linsong Chen, Ke-Jia Liu, Zheng Jiangsu Key Laboratory of Big Data Security & Intelligent Processing Nanjing University of Posts and Telecommunications Jiangsu Nanjing210023 China School of Computer Science Nanjing University of Posts and Telecommunications Jiangsu Nanjing210023 China State Key Laboratory of Novel Software Technology Nanjing University Jiangsu Nanjing210093 China

Directed graph is able to model asymmetric relationships between nodes and research on directed graph embedding is of great significance in downstream graph analysis and inference. Learning source and target embeddings of nodes separately to preserve edge asymmetry has become the dominant approach, but also poses challenge for learning representations of low or even zero in/out degree nodes that are ubiquitous in sparse graphs. In this paper, a collaborative bi-directional aggregation method (COBA) for directed graph embedding is proposed. Firstly, the source and target embeddings of the central node are learned by aggregating from the counterparts of the source and target neighbors, respectively;Secondly, the source/target embeddings of the zero in/out degree central nodes are enhanced by aggregating the counterparts of opposite-directional neighbors (i.e. target/source neighbors);Finally, source and target embeddings of the same node are correlated to achieve collaborative aggregation. Both the feasibility and rationality of the model are theoretically analyzed. Extensive experiments on real-world datasets demonstrate that COBA comprehensively outperforms state-of-the-art methods on multiple tasks and meanwhile validates the effectiveness of proposed aggregation strategies. © 2023, The Authors. All rights reserved.

关键词： Directed graphs

Multi-objective optimization for fuzzy workflow scheduling

学校读者我要写书评

暂无评论

Multi-objective optimization for fuzzy workflow scheduling

IEEE International Conference on Systems, Man and Cybernetics

作者： Jie Zhu Jing Zhang Chang Lu Haiping Huang Nanjing University of Posts and Telecommunications Nanjing China Jiangsu Key Laboratory of Big Data Security & Intelligent Processing Nanjing China Nanjing University of Posts & Telecommunications Nanjing China Jiangsu High Technology Research Key Laboratory for Wireless Sensor Networks Nanjing China Nanjing University of Aeronautics and Astronautics Nanjing China

ISBN: (纸本)9781665442084

A fuzzy workflow scheduling problem is investigated with fuzzy temporal parameters, such as the fuzzy task processing times, fuzzy data transmission times and the fuzzy due dates. In the considered problem, the resources are elastic cloud resources with multiple price structures. Due to the fuzzy temporal parameters, the deadline constraint cannot be ensured. Therefore, we formulate a triangle fuzzy number-based workflow scheduling problem model with the soft deadline constraint. A greedy local search method is proposed. The objectives are minimizing the total rental cost and the average dissatisfaction degree. The Pareto-based metrics are introduced to evaluate the performance of the compared algorithms. Experimental results show that the proposal is more effective and robust than the two existing algorithms.

关键词： Measurement Schedules Costs Processor scheduling Search methods Proposals data communication

Identifying outliers in astronomical images with unsupervised machine learning

学校读者我要写书评

暂无评论

arXiv

arXiv 2022年

作者： Han, Yang Zou, Zhiqiang Li, Nan Chen, Yanli School of Computer Science Nanjing University of Posts and Telecommunications Jiangsu Nanjing China Jiangsu Key Laboratory of Big Data Security and Intelligent Processing Jiangsu Nanjing China Key Laboratory of Optical Astronomy National Astronomical Observatories Chinese Academy of Sciences Beijing China University of Chinese Academy of Sciences Beijing China

Astronomical outliers, such as unusual, rare or unknown types of astronomical objects or phenomena, constantly lead to the discovery of genuinely unforeseen knowledge in astronomy. More unpredictable outliers will be uncovered in principle with the increment of the coverage and quality of upcoming survey data. However, it is a severe challenge to mine rare and unexpected targets from enormous data with human inspection due to a significant workload. Supervised learning is also unsuitable for this purpose since designing proper training sets for unanticipated signals is unworkable. Motivated by these challenges, we adopt unsupervised machine learning approaches to identify outliers in the data of galaxy images to explore the paths for detecting astronomical outliers. For comparison, we construct three methods, which are built upon the k-nearest neighbors (KNN), Convolutional Auto-Encoder (CAE) + KNN, and CAE + KNN + Attention Mechanism (attCAE KNN) separately. Testing sets are created based on the Galaxy Zoo image data published online to evaluate the performance of the above methods. Results show that attCAE KNN achieves the best recall (78%), which is 53% higher than the classical KNN method and 22% higher than CAE+KNN. The efficiency of attCAE KNN (10 minutes) is also superior to KNN (4 hours) and equal to CAE+KNN (10 minutes) for accomplishing the same task. Thus, we believe it is feasible to detect astronomical outliers in the data of galaxy images in an unsupervised manner. Next, we will apply attCAE KNN to available survey datasets to assess its applicability and reliability. Copyright © 2022, The Authors. All rights reserved.

关键词： k-nearest neighbors

Throughput Maximization for UAV-Enabled data Collection

学校读者我要写书评

暂无评论

Throughput Maximization for UAV-Enabled Data Collection

International Conference on Mobile Ad-hoc and Sensor Networks, MSN

作者： Junchao Gong Xiaojun Zhu Lijie Xu College of Computer Science and Technology Nanjing University of Aeronautics and Astronautics Nanjing China Jiangsu Key Laboratory of Big Data Security & Intelligent Processing Nanjing University of Posts and Telecommunications Nanjing China Collaborative Innovation Center of Novel Software Technology and Industrialization Nanjing China

ISBN: (数字)9781728199160

ISBN: (纸本)9781728192369

In this paper, we consider a scenario where an unmanned aerial vehicle (UAV) collects data from a set of nodes placed on a two-dimensional (2-D) plane. The UAV flies along a given line to collect data. To prevent transmission collisions, only one node is allowed to transmit data to the UAV at any time, and the number of nodes that can transmit data to the UAV cannot exceed a specified number to avoid frequent switches. The problem is to select a subset of nodes and schedule their transmission time to maximize the UAV's throughput. After formulating the problem, we find that it is difficult to solve due to mixture of real variables and integer variables. We then propose to decompose the problem into subproblems, and give a polynomial time algorithm to solve each subproblem to optimality. We then get an exact algorithm by solving all subproblems. Unfortunately, due to an exponential number of subproblems, the overall running time is exponential with respect to the input size. We then propose two polynomial-time suboptimal algorithms to solve the problem, both of which explore a polynomial number of subproblems. Simulations show that the suboptimal algorithms perform comparably with the exponential-time exact algorithm, while the running time is much smaller.

关键词： Schedules Heuristic algorithms Two dimensional displays data collection Throughput Unmanned aerial vehicles Sensors

Dlr: Adversarial Examples Detection and Label Recovery for Deep Neural Networks

学校读者我要写书评

暂无评论

SSRN 2024年

作者： Han, Keji Li, Yun Ge, Yao Wang, Ruchuan Nanjing University of Posts and Telecommunications Wenyuan Road 9 Nanjing210046 China Jiangsu Key Laboratory of Big Data Security & Intelligent Processing Wenyuan Road 9 Nanjing210046 China Jiangsu High Technology Research Key Laboratory for Wireless Sensor Networks Wenyuan Road 9 Nanjing210046 China

Deep neural networks (DNNs) are demonstrated to be vulnerable to the adversarial example crafted by the adversary to fool the target model. Adversarial training and adversarial example detection are two popular methods to mitigate the adversarial issue. The adversarial training makes the target model recognize the adversarial example correctly for the image classification task, while it does not generalize well. The adversarial detection generalizes well but cannot help the target model recognize the adversarial example. We first define the label recovery task to address the adversarial issue for DNNs. Then we propose a new generative classifier and apply it to the adversarial example label recovery task. The proposed method is named as Detection and Label Recovery (DLR) defense method, which consists of Detector and Recover. Detector feeds the legitimate and adversarial examples to the target model and Recover, respectively. The Recover explores the ground-truth label of the detected adversarial example. DLR combines the merits of adversarial training and adversarial example detection. Experimental results demonstrate that our method outperforms some state-of-the-art adversarial training methods. © 2024, The Authors. All rights reserved.

关键词： Deep neural networks

Pptcdr: Personalized Preference Transfer Cross-Domain Recommendation Based on Heterogeneous Graph Neural Networks

学校读者我要写书评

暂无评论

SSRN 2023年

作者： Zhu, Fan Xu, Xiaolong Jiangsu Key Laboratory of Big Data Security and Intelligent Processing Nanjing University of Posts and Telecommunications No. 9 Wenyuan Road Nanjing210023 China School of Computer Science Nanjing University of Posts and Telecommunications No. 9 Wenyuan Road Nanjing210023 China

Cross-domain recommendation methods aim to solve the problems of data sparsity and cold start by leveraging rich information from the source domain to enhance the recommendation accuracy of the target domain. Many methods ignore auxiliary information, which is often structured. Therefore, based on the above issue, we propose a method called PPTCDR (Personalized Preference Transfer Cross-Domain Recommendation) based on heterogeneous graph neural networks. PPTCDR introduces auxiliary information to enrich the information of the source and target domains, alleviates the problem of inaccurate embedding caused by data sparsity, and performs feature embedding of users and items separately in the source and target domains. In addition, for user preference transfer, in order to achieve personalization, a personalized bridging function is provided for each user. PPTCDR first constructs heterogeneous graphs separately in the source and target domains, and learns the latent embeddings of users and items using heterogeneous graph neural networks. Then, a meta-network is used to generate a personalized bridging function for each user. Through the personalized bridging function, the user's embedding in the source domain can be transformed to the target domain, and the transformed user embedding is used as the initial embedding of cold start users in the target domain. In this paper, we set up three cross-domain tasks on the Amazon dataset to verify and compare the performance of the experiment. PPTCDR achieves MAE and RMSE performances of 0.67 and 0.94 respectively, which are respectively improved by 23% and 18% compared to the state-of-the-art works EMCDR and PTUPCDR. © 2023, The Authors. All rights reserved.

关键词： Embeddings

Sindiff: Spoken-to-Sign Language Generation Based Transformer Diffusion Model

学校读者我要写书评

暂无评论

SSRN 2023年

作者： Liang, Wuyan Xu, Xiaolong Jiangsu Key Laboratory of Big Data Security & Intelligent Processing Nanjing University of Posts and Telecommunications No. 9 Wenyuan Road Nanjing210023 China School of Computer Science Nanjing University of Posts and Telecommunications No. 9 Wenyuan Road Nanjing210023 China

Sign language generation (SLG) is meaningful in facilitating the communication between deaf and healthy people. Recently, transformer-based diffusion models have been achieved success in human motion generation. However, it is still under-explored how to generate long-term sign language with diffusion model. Here, we propose a transformer-based diffusion framework, SinDiff, that explore all the advantages of transformer (i.e., dynamic attention and global context) for spoken-driven SLG. To meet the goal, we develop a linear biased multi-head self-attention with periodic position encoding to generalize longer-term sign language, and a linear biased cross-modal multi-head attention function to align the spoken and sign language modalities. Besides, a transformer-based spoken encoder is introduced to solve the limitations of sign data and learn the global spoken semantics. By experimental results on our testbed, we show that the word accuracy of the SinDiff for the back-translated spoken translation with ground truth spoken reach 21.02%, which has a 2.8% improvement over the 18.22% accuracy of the typical SLG method (GEN-OBT) on PHOENIX14T test set. And the generative visualization results further verify the feasibility of the proposed method. © 2023, The Authors. All rights reserved.

关键词： Semantics

Universal deep network for steganalysis of color image based on channel representation

学校读者我要写书评

暂无评论

arXiv

arXiv 2021年

作者： Wei, Kangkang Luo, Weiqi Tan, Shunquan Huang, Jiwu The Guangdong Key Lab of Information Security Technology School of Computer Science and Engineering Sun Yat-Sen University Guangzhou510006 China The College of Computer Science and Software Engineering Shenzhen University Shenzhen518060 China The Guangdong Key Laboratory of Intelligent Information Processing Shenzhen Key Laboratory of Media Security National Engineering Laboratory for Big Data System Computing Technology Shenzhen University Shenzhen518060 China

Up to now, most existing steganalytic methods are designed for grayscale images, and they are not suitable for color images that are widely used in current social networks. In this paper, we design a universal color image steganalysis network (called UCNet) in spatial and JPEG domains. The proposed method includes preprocessing, convolutional, and classification modules. To preserve the steganographic artifacts in each color channel, in preprocessing module, we firstly separate the input image into three channels according to the corresponding embedding spaces (i.e. RGB for spatial steganography and YCbCr for JPEG steganography), and then extract the image residuals with 62 fixed high-pass filters, finally concatenate all truncated residuals for subsequent analysis rather than adding them together with normal convolution like existing CNN-based steganalyzers. To accelerate the network convergence and effectively reduce the number of parameters, in convolutional module, we carefully design three types of layers with different shortcut connections and group convolution structures to further learn high-level steganalytic features. In classification module, we employ a global average pooling and fully connected layer for classification. We conduct extensive experiments on ALASKA II to demonstrate that the proposed method can achieve state-of-the-art results compared with the modern CNN-based steganalyzers (e.g., SRNet and J-YeNet) in both spatial and JPEG domains, while keeping relatively few memory requirements and training time. Furthermore, we also provide necessary descriptions and many ablation experiments to verify the rationality of the network design. Copyright © 2021, The Authors. All rights reserved.

关键词： Steganography