检索结果-内蒙古大学图书馆

Logiformer: A Two-Branch Graph Transformer Network for Interpretable Logical Reasoning

学校读者我要写书评

暂无评论

arxiv 2022年

作者： Xu, Fangzhi Liu, Jun Lin, Qika Pan, Yudai Zhang, Lingling School of Computer Science and Technology Xi'an Jiaotong University Xi'an China Shaanxi Province Key Laboratory of Satellite and Terrestrial Network Tech. R&d National Engineering Lab for Big Data Analytics Xi'an China

Machine reading comprehension has aroused wide concerns, since it explores the potential of model for text understanding. To further equip the machine with the reasoning capability, the challenging task of logical reasoning is proposed. Previous works on logical reasoning have proposed some strategies to extract the logical units from different aspects. However, there still remains a challenge to model the long distance dependency among the logical units. Also, it is demanding to uncover the logical structures of the text and further fuse the discrete logic to the continuous text embedding. To tackle the above issues, we propose an end-to-end model Logiformer which utilizes a two-branch graph transformer network for logical reasoning of text. Firstly, we introduce different extraction strategies to split the text into two sets of logical units, and construct the logical graph and the syntax graph respectively. The logical graph models the causal relations for the logical branch while the syntax graph captures the co-occurrence relations for the syntax branch. Secondly, to model the long distance dependency, the node sequence from each graph is fed into the fully connected graph transformer structures. The two adjacent matrices are viewed as the attention biases for the graph transformer layers, which map the discrete logical structures to the continuous text embedding space. Thirdly, a dynamic gate mechanism and a question-aware self-attention module are introduced before the answer prediction to update the features. The reasoning process provides the interpretability by employing the logical units, which are consistent with human cognition. The experimental results show the superiority of our model, which outperforms the state-of-the-art single model on two logical reasoning benchmarks. © 2022, CC BY.

关键词： Embeddings

Distilled Transformers with Locally Enhanced Global Representations for Face Forgery Detection

学校读者我要写书评

暂无评论

arxiv 2024年

作者： Zhang, Yaning Li, Qiufu Yu, Zitong Shen, Linlin Computer Vision Institute College of Computer Science and Software Engineering Shenzhen University Guangdong Shenzhen518060 China National Engineering Laboratory for Big Data System Computing Technology Shenzhen University Guangdong Shenzhen518060 China Shenzhen Institute of Artificial Intelligence and Robotics for Society Guangdong Shenzhen518129 China Guangdong Key Laboratory of Intelligent Information Processing Shenzhen University Guangdong Shenzhen518060 China School of Computing and Information Technology Great Bay University Guangdong Dongguan523000 China

Face forgery detection (FFD) is devoted to detecting the authenticity of face images. Although current CNN-based works achieve outstanding performance in FFD, they are susceptible to capturing local forgery patterns generated by various manipulation methods. Though transformer-based detectors exhibit improvements in modeling global dependencies, they are not good at exploring local forgery artifacts. Hybrid transformer-based networks are designed to capture local and global manipulated traces, but they tend to suffer from the attention collapse issue as the transformer block goes deeper. Besides, soft labels are rarely available. In this paper, we propose a distilled transformer network (DTN) to capture both rich local and global forgery traces and learn general and common representations for different forgery faces. Specifically, we design a mixture of expert (MoE) module to mine various robust forgery embeddings. Moreover, a locally-enhanced vision transformer (LEVT) module is proposed to learn locally-enhanced global representations. We design a lightweight multi-attention scaling (MAS) module to avoid attention collapse, which can be plugged and played in any transformer-based models with only a slight increase in computational costs. In addition, we propose a deepfake self-distillation (DSD) scheme to provide the model with abundant soft label information. Extensive experiments show that the proposed method surpasses the state of the arts on five deepfake datasets. Copyright © 2024, The Authors. All rights reserved.

关键词： Face recognition

AdaSemiCD: An Adaptive Semi-supervised Change Detection Method Based on Pseudo Label Evaluation

学校读者我要写书评

暂无评论

arxiv 2024年

作者： Ran, Lingyan Wen, Dongcheng Zhuo, Tao Zhang, Shizhou Zhang, xiuwei Zhang, Yanning School of Computer Science Shaanxi Provincial Key Laboratory of Speech and Image Information Processing The National Engineering Laboratory for Integrated Aerospace-GroundOcean Big Data Application Technology The Ningbo Institute of Northwestern Polytechnical University Northwestern Polytechnical University Xi’an710072 China College of Information Engineering Northwest A&F University Yangling712100 China

Change detection(CD) is an essential field in remote sensing, with a primary focus on identifying areas of change in bitemporal image pairs captured at varying intervals of the same region. The data annotation process for CD tasks is both time-consuming and labor-intensive. To better utilize the scarce labeled data and abundant unlabeled data, we introduce an adaptive semi-supervised learning method, AdaSemiCD, to improve pseudo-label usage and optimize the training process. Initially, due to the extreme class imbalance inherent in CD, the model is more inclined to focus on the background class, and it is easy to confuse the boundary of the target object. Considering these two points, we develop a measurable evaluation metric for pseudo-labels that enhances the representation of information entropy by class rebalancing and amplification of ambiguous areas, assigning greater weights to prospective change objects. Subsequently, to enhance the reliability of samplewise pseudo-labels, we introduce the AdaFusion module, to dynamically identifying the most uncertain region and substituting it with more trustworthy content. Lastly, to ensure better training stability, we introduce the AdaEMA module, which updates the teacher model using only batches of trusted samples. Experimental results on ten public CD datasets validate the efficacy and generalizability of our proposed adaptive training framework. Copyright © 2024, The Authors. All rights reserved.

关键词： Self-supervised learning

New Approach for Efficient Malicious Multiparty Private Set Intersection

学校读者我要写书评

暂无评论

SSRN

SSRN 2023年

作者： Lv, Siyi Wei, Yu Jia, Jingyu Li, xinhao Liu, Zheli Chen, xiaofeng Guo, Liang College of Cyber Science The College of Computer Science Key Laboratory of Data and Intelligent System Security Ministry of Education Nankai University Tianjin China Department of Computer Science Purdue University United States School of Cyber Engineering Xidian University Xi’an China Huawei Technology Co. Ltd Shenzhen China

The Multiparty Private Set Intersection (MPSI) protocol allows untrusted parties to compute the intersection of their sets without revealing any additional information. However, existing solutions face challenges in terms of security and efficiency. Some schemes are secure only against semi-honest adversaries and cannot effectively defend against malicious adversaries. Other schemes, while capable of withstanding malicious adversaries, exhibit efficiency problems, especially as the number of corrupt participants increases. A practical solution must maintain high security and efficiency, even when handling large-scale data and a majority of corrupted *** paper introduces a novel scheme named TH-PSI, built upon a new star-shaped topological network communication framework based on the trusted execution environment. TH-PSI has been rigorously proven to be secure within the standard UC framework against malicious adversaries. To demonstrate the efficiency of TH-PSI, we implement the TH-PSI scheme and tested it in both LAN and WAN settings, involving up to 30 participants and datasets as large as $2^{22}$. We compare TH-PSI with state-of-the-art schemes, ENOC21 and NTY21. Our experiments show that TH-PSI is an efficient solution capable of delivering results within a matter of seconds, even when dealing with data at the million-level and a majority of corrupted participants. © 2023, The Authors. All rights reserved.

关键词： Efficiency

FRSM: A Novel Fault-Tolerant Approach for Redundant-Path-Enabled Service Migration in Mobile Edge Computing 1

学校读者我要写书评

暂无评论

29th International Conference on Web Services, ICWS 2022 Held as Part of the Services Conference Federation, SCF 2022

作者： Zhao, Jiale Dai, Mengxuan xia, Yunni Ma, Yong He, Meibin Peng, Kai Li, Jianqi Li, Fan Fu, xiaodong School of Computer Chongqing University Chongqing400030 China School of Computer and Information Engineering Jiangxi Normal University Nanchang330022 China Department of Big Data and AI of Jiangxi Telecom Nanchang330029 China Engineering College Huaqiao University Quanzhou362021 China Global Energy Interconnection Research Institute Co. Ltd. Beijing102209 China Key Laboratory of Fundamental Synthetic Vision Graphics and Image Science for National Defense Sichuan University Chengdu610065 China Data Science Research Center Faculty of Science Kunming University of Science and Technology Kunming650031 China

ISBN: (数字)9783031235795

ISBN: (纸本)9783031235788

Mobile Edge Computing (MEC) provides users with low-latency, highly responsive services by deploying Edge Servers (ESs) near applications. MEC allows any edge-hosted application or service to be migrated between different edge resource providers without being locked into a single provider. Nevertheless, due to its complexity and dynamics, the real edge computing environment is prone to errors and failures, reducing the reliability of edge service migration. This paper proposes a novel fault-tolerant method for redundant path service migration. The method utilizes sliding-window-based model and identifies a set of service migration paths, enabling the evaluation of the time-varying failure rate of ESs. The method combines resubmission and replication mechanisms and decides the edge service migration scheme by selecting multiple redundant migration paths. We also conduct extensive simulations and show that our proposed method outperforms traditional solutions in several metrics. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Mobile edge computing

A Hybrid Prediction Method Based on MODWT-EMD for Time Series in Wireless Sensor Networks

学校读者我要写书评

暂无评论

A Hybrid Prediction Method Based on MODWT-EMD for Time Serie...

International Conference on intelligent System and Knowledge Engineering, ISKE

作者： Zhongmin Wang Jingwen Jia Cong Gao Yanping Chen Hong xia Xi’an Key Laboratory of Big Data and Intelligent Computing Xi’an University of Posts and Telecommunications Xi’an China School of Computer Science & Technology Xi’an University of Posts and Telecommunications Xi’an China Shaanxi Key Laboratory of Network Data Analysis and Intelligent Processing Xi’an University of Posts and Telecommunications Xi’an China

ISBN: (纸本)9781665405546

Time series is the main form of sensor data. Time series prediction is conducive to reduce the energy consumption of sensor nodes and increase the service life. As time cost and computation cost of the existing hybrid prediction methods for time series are not good, this paper proposes a hybrid prediction method based on MODWT-EMD for time series. Firstly, the original time series is decomposed into multiple frequency components by Maximal Overlapping Discrete Wavelet Transform (MODWT). Secondly, the multiple frequency components are decomposed successively by Empirical Mode Decomposition (EMD) to obtain IMF components and residuals. IMFs and residuals contain the feature information of time series. Then, the importance of the extracted feature information is scored and sorted by random forest. We select feature information with a higher score. Finally, the selected feature information is input into Bi-GRU for training. We predicted the original time series by the trained Bi-GRU and obtained the prediction results. The experimental results show that the method proposed in this paper can predict the time series effectively.

关键词： Training Wireless communication Wireless sensor networks Time-frequency analysis Costs Time series analysis Predictive models

Video Smoke Detection Based on Motion Analysis and Fused Temporal-Spatial Model

学校读者我要写书评

暂无评论

SSRN

SSRN 2022年

作者： Jia, Yang Mao, Zixu Zhang, xinmeng Chen, Weiguang Wang, Ben Wang, Gang Zhang, Qixing School of Computer Science Xi'an University of Posts and Telecommunications Xi'An710121 China Shaanxi Key Laboratory of Network Data Intelligent Processing Xi'an University of Posts and Telecommunications Xi'An710121 China XNIC262 Xi'An710121 China School of Information Science and Technology University of Science and Technology of China Hefei230026 China State Key Laboratory of Fire Science University of Science and Technology of China Hefei230026 China

Researchers have proposed various video smoke detection methods based on deep learning. However, most methods are based on two-dimensional image analysis, which hasn't effectively used the motion information of smoke. Therefore, an improved two-stream CNN based smoke detection method is proposed. Firstly, we use sparse sampling to extract sampled frames and capture the suspected area of the sampled frame with background modeling. Then we extract the dense optical flow of consecutive frames of the sampled frames and merge them into a stacked optical flow sequence. A sampled frame and a stacked optical flow sequence are fed into the Two-Stream CNN to obtain video-level recognition results. The performance of 97.32% accuracy, and 4.07% false alarm rate can be achieved, and the effectiveness of the method has been proved through comparative experiments. Combined with the explicitly expressed motion information, we can get a better detection performance. © 2022, The Authors. All rights reserved.

关键词： Motion analysis

An Efficient Image Captioning Method Based on Generative Adversarial Networks 21

学校读者我要写书评

暂无评论

An Efficient Image Captioning Method Based on Generative Adv...

Proceedings of the 2021 4th International Conference on Artificial Intelligence and Pattern Recognition

作者： Lin Zhenxian Feng Feirong Yang xiaobao Ding Chen School of Science Xi'an University of Posts and Telecommunications China School of Telecommunication and Information Engineering Xi'an University of Posts and Telecommunications China School of Computer Science and Technology Northwestern Polytechnical UniversityChina. China and School of Computer Science and TechnologyXi'an University of Posts and Telecommunications China School of Computer Science and Technology Xi'an University of Posts and TelecommunicationsChina China and Shaanxi Key Laboratory of Network Data Analysis and Intelligent ProcessingXi'an University of Posts and Telecommunications China

ISBN: (纸本)9781450384087

Image captioning is usually based on the mainstream encoder-decoder framework, the accuracy of the encoders and the generation capability of the decoders directly affect the quality of image captions. But no matter how the encoders and the decoders are combined, the generated captions will only imitate the ground-truth captions, and its diversity is suppressed. To generate more accurate and diverse captions while exploring the impact of different encoder-decoder frameworks for the quality of the generated captions, we conducted an exploration of an efficient image captioning method based on GAN. Specifically, we use ResNet152, EfficientNet-B0, and EfficientNet-B2 as the encoders to extract image features, GRU or Transformer as the decoders to generate captions, respectively. Based on MS COCO2014 dataset, a large number of experiment results show that the EfficientNet-B2+GRU framework has best performance, training time is reduced, BLUE-4 score increases by 0.6, parameters reduce by 0.66 million, which can be used as an efficient encoder-decoder framework for generating diverse captions.

关键词： Decoder Encoder Generative Adversarial Networks Image captioning

GPTR: Gestalt-Perception Transformer for Diagram Object Detection

学校读者我要写书评

暂无评论

arxiv 2022年

作者： Hu, xin Zhang, Lingling Liu, Jun Fan, Jinfu You, Yang Wu, Yaqiang Shaanxi Provincial Key Laboratory of Big Data Knowledge Engineering School of Computer Science and Technology Xi’an Jiaotong University China National Engineering Lab for Big Data Analytics Xi’an Jiaotong University China Department of Control Science and Engineering Tongji University Shanghai China Department of Computer Science National University of Singapore Singapore Lenovo Research Beijing China

Diagram object detection is the key basis of practical applications such as textbook question answering. Because the diagram mainly consists of simple lines and color blocks, its visual features are sparser than those of natural images. In addition, diagrams usually express diverse knowledge, in which there are many low-frequency object categories in diagrams. These lead to the fact that traditional data-driven detection model is not suitable for diagrams. In this work, we propose a gestalt-perception transformer model for diagram object detection, which is based on an encoder-decoder architecture. Gestalt perception contains a series of laws to explain human perception, that the human visual system tends to perceive patches in an image that are similar, close or connected without abrupt directional changes as a perceptual whole object. Inspired by these thoughts, we build a gestalt-perception graph in transformer encoder, which is composed of diagram patches as nodes and the relationships between patches as edges. This graph aims to group these patches into objects via laws of similarity, proximity, and smoothness implied in these edges, so that the meaningful objects can be effectively detected. The experimental results demonstrate that the proposed GPTR achieves the best results in the diagram object detection task. Our model also obtains comparable results over the competitors in natural image object detection. Copyright © 2022, The Authors. All rights reserved.

关键词： Object detection