检索结果-内蒙古大学图书馆

Deep learning-based hyperspectral and multispectral fusion techniques: Review, optimization, and perspectives

学校读者我要写书评

暂无评论

Information Fusion 2025年 124卷

作者： Sen Luo Yurong Qian Lu Bai Yingying Fan Yuanxu Wang WeiQuan Kong Xinjiang Engineering Research Center of Big Data and Intelligent Software School of Software Xinjiang University Urumqi 830091 China Joint International Research Laboratory of Silk Road Multilingual Cognitive Computing Xinjiang University Urumqi 830046 China School of Computer Science and Technology Xinjiang University Urumqi 830046 China

Hyperspectral and Multispectral Image Fusion (HMIF) aim to acquire high spatial and spectral resolution images, which overcomes the inherent limitations of a single imaging device and effectively extends the applicability of spectral images on various downstream tasks. In recent years, Deep Learning (DL)-based image fusion technology breaks through the many limitations of traditional methods, rapidly develops a variety of neural network architectures and improvement strategies, and gradually overcomes a series of practical difficulties. However, a comprehensive overview of these deep learning architectures and various improvement techniques is still lacking. In this paper, we present a comprehensive review of deep learning techniques on HMIF, organize the fusion paradigms according to different researchers’ understanding and hypotheses of the problem, and sort out the common optimization strategies of neural network models, such as attention mechanism, multi-feature learning, combined with the variational and unmixing models of deep learning. In addition, we also provide a detailed compendium of the practical challenges and corresponding solution strategies for the first time, which involves the spatial-temporal misregistration, generalization enhancement and high-magnification fusion problems. Finally, we sort out and categorize the commonly used evaluation metrics for fused datasets, and discuss the potential research directions in HMIF. Through these works, we hope to provide a quick and clear reference for aspiring researchers in this area.

关键词：

CoSimGNN: Towards Large-scale Graph Similarity Computation

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Xu, Haoyan Chen, Runjian Wang, Yueyang Duan, Ziheng Feng, Jie School of Big Data and Software Engineering Chongqing University Chongqing401331 China College of Energy Engineering Zhejiang University Zhejiang310027 China

The ability to compute similarity scores between graphs based on metrics such as Graph Edit Distance (GED) is important in many real-world applications. Computing exact GED values is typically an NP-hard problem and traditional algorithms usually achieve an unsatisfactory trade-off between accuracy and efficiency. Recently, Graph Neural Networks (GNNs) provide a data-driven solution for this task, which is more efficient while maintaining prediction accuracy in small graph (around 10 nodes per graph) similarity computation. Existing GNN-based methods, which either respectively embed two graphs (lack of low-level cross-graph interactions) or deploy cross-graph interactions for whole graph pairs (redundant and time-consuming), are still not able to achieve competitive results when the number of nodes in graphs increases. In this paper, we focus on similarity computation for large-scale graphs and propose the "embedding-coarsening-matching" framework COSIMGNN, which first embeds and coarsens large graphs with adaptive pooling operation and then deploys fine-grained interactions on the coarsened graphs for final similarity scores. Furthermore, we create several synthetic datasets which provide new benchmarks for graph similarity computation. Detailed experiments on both synthetic and real-world datasets have been conducted and COSIMGNN achieves the best performance while the inference time is at most 1/3 of that of previous state-of-the-art. Copyright © 2020, The Authors. All rights reserved.

关键词： Graph neural networks

An Ensemble of Random Decision Trees with Personalized Privacy Preservation in Edge-Cloud Computing 13

学校读者我要写书评

暂无评论

An Ensemble of Random Decision Trees with Personalized Priva...

2020 IEEE Congress on Cybermatics: 13th IEEE International Conferences on Internet of Things, iThings 2020, 16th IEEE International Conference on Green Computing and Communications, GreenCom 2020, 13th IEEE International Conference on Cyber, Physical and Social Computing, CPSCom 2020 and 6th IEEE International Conference on Smart data, Smartdata 2020

作者： Wu, Xiaotong Xu, Xiaolong Dai, Fei Gao, Jiaquan Ji, Genlin Qi, Lianyong School of Computer and Electronic Information Nanjing Normal University Nanjing China School of Computer and Software Nanjing University of Information Science Technology Nanjing China School of Big Data and Intelligence Engineering Southwest Forestry University Kunming China School of Information Science and Engineering Qufu Normal University Qufu China

ISBN: (纸本)9781728176475

Edge-cloud computing (ECC) is an emerging computing paradigm, which offers a great opportunity to implement data mining-based services and applications for a large number of devices and sensors in Internet of Things. However, data privacy has been one of the most concerned problems for all the participants. In this paper, we propose a framework of privacy-preserving data mining based on private random decision trees in ECC. The framework not only gives the strong privacy guarantee, but also provides the efficient data utility. At first, we design a differential privacy-based framework to implement private random decision trees in ECC. Then, we present the concrete algorithms and the corresponding task that each participant needs to undertake. Next, we analyze the key factors to influence privacy and utility and give further improvement to increase the utility with strong privacy preservation. Finally, extensive experiments demonstrate the good performance of our designed framework. © 2020 IEEE.

关键词： data mining

Sentence Pair Similarity Modeling Based on Weighted Interaction of Multi-semantic Embedding Matrix

学校读者我要写书评

暂无评论

Sentence Pair Similarity Modeling Based on Weighted Interact...

International Conference on Tools for Artificial Intelligence (ICTAI)

作者： Junyu Chen Xiaohong Zhu Jun Sang Lu Gong College of Computer Science Chongqing University Chongqing China School of Big Data & Software Engineering Chongqing University Chongqing China

ISBN: (数字)9781728192284

ISBN: (纸本)9781728185361

In this paper, we focus on measuring the similarity of sentence pair. Noting that a single sentence vector may lose fine-grained semantic information which is important for sentence matching, we propose an embedding matrix to calculate a multi-granularity similarity matrix and find the true semantic alignment of two sentences. We also propose a semantic importance calculation and semantic decomposition that are simple but effective. The proposed model does not require any sparse features or external resources such as WordNet. Compared with other state-of-the-art models, we successfully train in a short time and achieve competitive results on similarity measurement and paraphrase identification tasks. Experiments and visual analysis show the good performance and interpretability of the model.

关键词： Weight measurement Training Visualization Analytical models Semantics Matrix decomposition Sparse matrices

Low-Complexity Channel Estimation for Massive MIMO Systems with Decentralized Baseband Processing

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Xu, Yanqing Wang, Bo Song, Enbin Shi, Qingjiang Chang, Tsung-Hui The School of Science and Engineering The Chinese University of Hong Kong Shenzhen China The Shenzhen Research Institute of Big Data Shenzhen518172 China The Wireless Network RAN Algorithm Department Xi’an Huawei Technologies Co. Ltd. Xi’an710000 China The College of Mathematics School of Aeronautics and Astronautics Sichuan University Sichuan Chengdu610064 China The School of Software Engineering Tongji University Shanghai201804 China

The traditional centralized baseband processing architecture is faced with the bottlenecks of high computation complexity and excessive fronthaul communication, especially when the number of antennas at the base station (BS) is large. To cope with these two challenges, the decentralized baseband processing (DPB) architecture has been proposed, where the BS antennas are partitioned into multiple clusters and each is connected to a local baseband unit (BBU). In this paper, we are interested in the low-complexity distributed channel estimation (CE) method under such DBP architecture, which is rarely studied in the literature. The aim is to devise distributed CE algorithms that can perform as well as the centralized scheme but with a small inter-BBU communication cost. Specifically, based on the low-complexity diagonal minimum mean square error channel estimator, we propose two distributed CE algorithms, namely the aggregate-then-estimate algorithm and the estimate-then-aggregate algorithm. In contrast to the existing distributed CE algorithm which requires iterative information exchanges among the nodes, our algorithms only require one roundtrip communication among BBUs. Extensive experiment results are presented to demonstrate the advantages of the proposed distributed CE algorithms in terms of estimation accuracy, inter-BBU communication cost and computation complexity. Copyright © 2022, The Authors. All rights reserved.

关键词： Channel estimation

A cross Transformer for image denoising

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Tian, Chunwei Zheng, Menghua Zuo, Wangmeng Zhang, Shichao Zhang, Yanning Lin, Chia-Wen School of Software Northwestern Polytechnical University Shaanxi Xi’an710129 China National Engineering Laboratory for Integrated Aero-Space-Ground-Ocean Big Data Application Technology Shaanxi Xi’an710129 China School of Computer Science and Technology Harbin Institute of Technology Heilongjiang Harbin150001 China Peng Cheng Laboratory Shenzhen518055 China School of Computer Science and Engineering Central South University Hunan Changsha410083 China School of Computer Science Northwestern Polytechnical University Shaanxi Xi’an710129 China Department of Electrical Engineering The Institute of Communications Engineering National Tsing Hua University Hsinchu30013 Taiwan

Deep convolutional neural networks (CNNs) depend on feedforward and feedback ways to obtain good performance in image denoising. However, how to obtain effective structural information via CNNs to efficiently represent given noisy images is key for complex scenes. In this paper, we propose a cross Transformer denoising CNN (CTNet) with a serial block (SB), a parallel block (PB), and a residual block (RB) to obtain clean images for complex scenes. A SB uses an enhanced residual architecture to deeply search structural information for image denoising. To avoid loss of key information, PB uses three heterogeneous networks to implement multiple interactions of multi-level features to broadly search for extra information for improving the adaptability of an obtained denoiser for complex scenes. Also, to improve denoising performance, Transformer mechanisms are embedded into the SB and PB to extract complementary salient features for effectively removing noise in terms of pixel relations. Finally, a RB is applied to acquire clean images. Experiments illustrate that our CTNet is superior to some popular denoising methods in terms of real and synthetic image denoising. It is suitable to mobile digital devices, i.e., phones. Codes can be obtained at https://***/hellloxiaotian/CTNet. Copyright © 2023, The Authors. All rights reserved.

关键词： Image denoising

FLFE: A communication-efficient and privacy-preserving federated feature engineering framework

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Fang, Pei Cai, Zhendong Chen, Hui Shi, QingJiang School of Software Engineering Tongji University Shanghai201804 China Shenzhen Research Institution of Big Data Shenzhen518172 China

Feature engineering is the process of using domain knowledge to extract features from raw data via data mining techniques and is a key step to improve the performance of machine learning algorithms. In the multi-party feature engineering scenario (features are stored in many different IoT devices), direct and unlimited multivariate feature transformations will quickly exhaust memory, power, and bandwidth of devices, not to mention the security of information threatened. Given this, we present a framework called FLFE to conduct privacy-preserving and communication-preserving multi-party feature transformations. The framework pre-learns the pattern of the feature to directly judge the usefulness of the transformation on a feature. Explored the new useful feature, the framework forsakes the encryption-based algorithm for the well-designed feature exchange mechanism, which largely decreases the communication overhead under the premise of confidentiality. We made experiments on datasets of both open-sourced and real-world thus validating the comparable effectiveness of FLFE to evaluation-based approaches, along with the far more superior eicacy. Copyright © 2020, The Authors. All rights reserved.

关键词： Machine learning

An identity-based anonymous signcryption for vehicular ad hoc networks 2nd

学校读者我要写书评

暂无评论

An identity-based anonymous signcryption for vehicular ad ho...

2nd International Cognitive Cities Conference, IC3 2019

作者： Lu, Yong Yang, Huijie School of Management Science and Engineering Nanjing University of Information Science and Technology Nanjing China Guizhou Provincial Key Laboratory of Public Big Data Guizhou University Guiyang China School of Computer and Software Nanjing University of Information Science and Technology Nanjing China Guangdong Provincial Key Laboratory of Data Security and Privacy Protection Guangzhou China

ISBN: (纸本)9789811561122

The vehicular ad hoc networks (VANETs) can share information of vehicles’ location, speed and route within a certain range by interconnecting vehicles, roadside units (RSU) and the cloud. As people’s daily traffic tools, vehicles reflect users’ daily routines and other privacy. Therefore, privacy protection is very critical in VANETs. However, in the existing schemes, a user’s temporary identity (TID) is generated by the user himself, which is inconvenient to be managed and can be easily forged. The reputation values of users also need to be considered. In this paper, an identity based signcryption scheme for VANETs is proposed. In the novel scheme, TID is generated by key generation center (KCG) and a personal reputation system is produced. The security analysis of the scheme shows that our scheme is practical. © Springer Nature Singapore Pte Ltd 2020.

关键词： Vehicular ad hoc networks

Group-wise Inhibition based Feature Regularization for Robust Classification

学校读者我要写书评

暂无评论

Group-wise Inhibition based Feature Regularization for Robus...

International Conference on Computer Vision (ICCV)

作者： Haozhe Liu Haoqian Wu Weicheng Xie Feng Liu Linlin Shen Computer Vision Institute College of Computer Science and Software Engineering SZU Branch Shenzhen Institute of Artificial Intelligence and Robotics for Society National Engineering Laboratory for Big Data System Computing Technology Guangdong Key Laboratory of Intelligent Information Processing Shenzhen University Shenzhen China

The convolutional neural network (CNN) is vulnerable to degraded images with even very small variations (e.g. corrupted and adversarial samples). One of the possible reasons is that CNN pays more attention to the most discriminative regions, but ignores the auxiliary features when learning, leading to the lack of feature diversity for final judgment. In our method, we propose to dynamically suppress significant activation values of CNN by group-wise inhibition, but not fixedly or randomly handle them when training. The feature maps with different activation distribution are then processed separately to take the feature independence into account. CNN is finally guided to learn richer discriminative features hierarchically for robust classification according to the proposed regularization. Our method is comprehensively evaluated under multiple settings, including classification against corruptions, adversarial attacks and low data regime. Extensive experimental results show that the proposed method can achieve significant improvements in terms of both robustness and generalization performances, when compared with the state-of-the-art methods. Code is available at https://***/LinusWu/TENET_Training.

关键词： Training Computer vision Codes Heuristic algorithms Robustness Classification algorithms Convolutional neural networks