检索结果-内蒙古大学图书馆

Rethinking Super-Resolution as Text-Guided Details Generation

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Ma, Chenxi Yan, Bo Lin, Qing Tan, Weimin Chen, Siming School of Computer Science Shanghai Key Laboratory of Intelligent Information Processing Shanghai Collaborative Innovation Center of Intelligent Visual Computing Fudan University China School of Data Science Fudan University China

Deep neural networks have greatly promoted the performance of single image super-resolution (SISR). Conventional methods still resort to restoring the single high-resolution (HR) solution only based on the input of image modality. However, the image-level information is insufficient to predict adequate details and photo-realistic visual quality facing large upscaling factors (×8, ×16). In this paper, we propose a new perspective that regards the SISR as a semantic image detail enhancement problem to generate semantically reasonable HR image that are faithful to the ground truth. To enhance the semantic accuracy and the visual quality of the reconstructed image, we explore the multi-modal fusion learning in SISR by proposing a Text-Guided Super-Resolution (TGSR) framework, which can effectively utilize the information from the text and image modalities. Different from existing methods, the proposed TGSR could generate HR image details that match the text descriptions through a coarse-to-fine process. Extensive experiments and ablation studies demonstrate the effect of the TGSR, which exploits the text reference to recover realistic images. © 2022, CC BY.

关键词： Deep neural networks

Enhancing Security and Privacy in Federated Learning using Low-Dimensional Update Representation and Proximity-Based Defense

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Li, Wenjie Fan, Kai Zhang, Jingyuan Li, Hui Lim, Wei Yang Bryan Yang, Qiang The State Key Laboratory of Integrated Service Networks Xidian University Xi’an710071 China The College of Computing and Data Science Nanyang Technological University 639798 Singapore The Department of Computer Science and Engineering Hong Kong University of Science and Technology 999077 Hong Kong

Federated Learning (FL) is a promising privacy-preserving machine learning paradigm that allows data owners to collaboratively train models while keeping their data localized. Despite its potential, FL faces challenges related to the trustworthiness of both clients and servers, particularly against curious or malicious adversaries. In this paper, we introduce a novel framework named Federated Learning with Low-Dimensional Update Representation and Proximity-Based defense (FLURP), designed to address privacy preservation and resistance to Byzantine attacks in distributed learning environments. FLURP employs LinfSample method, enabling clients to compute the l∞ norm across sliding windows of updates, resulting in a Low-Dimensional Update Representation (LUR). Calculating the shared distance matrix among LURs, rather than updates, significantly reduces the overhead of Secure MultiParty Computation (SMPC) by three orders of magnitude while effectively distinguishing between benign and poisoned updates. Additionally, FLURP integrates a privacy-preserving proximity-based defense mechanism utilizing optimized SMPC protocols to minimize communication rounds. Our experiments demonstrate FLURP’s effectiveness in countering Byzantine adversaries with low communication and runtime overhead. FLURP offers a scalable framework for secure and reliable FL in distributed environments, facilitating its application in scenarios requiring robust data management and security. Copyright © 2024, The Authors. All rights reserved.

关键词： Federated learning

A3-CodGen: A Repository-Level Code Generation Framework for Code Reuse with Local-Aware, Global-Aware, and Third-Party-Library-Aware

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Liao, Dianshu Pan, Shidong Sun, Xiaoyu Ren, Xiaoxue Huang, Qing Xing, Zhenchang Jin, Huan Li, Qinying School of Computing Australian National University Australia School of Computer Information Engineering Jiangxi Normal University China CSIRO’s Data61 Australia State Key Laboratory of Blockchain and Data Security Zhejiang University China Institute of Blockchain and Data Security China Jiangxi University of Technology China

LLM-based code generation tools are essential to help developers in the software development process. Existing tools often disconnect with the working context, i.e., the code repository, causing the generated code to be not similar to human developers. In this paper, we propose a novel code generation framework, dubbed A3-CodGen, to harness information within the code repository to generate code with fewer potential logical errors, code redundancy, and library-induced compatibility issues. We identify three types of representative information for the code repository: local-aware information from the current code file, global-aware information from other code files, and third-party-library information. Results demonstrate that by adopting the A3-CodGen framework, we successfully extract, fuse, and feed code repository information into the LLM, generating more accurate, efficient, and highly reusable code. The effectiveness of our framework is further underscored by generating code with a higher reuse rate, compared to human developers. This research contributes significantly to the field of code generation, providing developers with a more powerful tool to address the evolving demands in software development in practice. Copyright © 2023, The Authors. All rights reserved.

关键词： Software design

Deep Kernel Embedded Clustering Network

学校读者我要写书评

暂无评论

SSRN

SSRN 2024年

作者： Ren, Lina Huang, Ruizhang Chen, Yanping Lin, Chuan Qin, Yongbin State Key Laboratory of Public Big Data Text Computing and Cognitive Intelligence Engineering Research Center of National Education Ministry College of Computer Science and Technology Guizhou University Guiyang550025 China Department of Information Engineering Guizhou Light Industry Technical College Guiyang550025 China

In this paper, we propose a deep kernel embedded clustering network, namely DKEC, which learns data partitions with kernelized semantic embeddings of data samples via a self-supervised deep neural network. A kernelized semantic embedding learner is designed to learn high quality data representations, which not only capture the data semantic characteristics but also show good structural separable characteristics. Other than directly apply traditional predefined kernel mapping methods, the kernelized semantic embedding learner, investigated by a deep variational Gaussian process, is able to learn the kernel mapping according to the nature of various datasets automatically, while maintain reasonable computational complexity. The DKEC, which is an end-to-end deep clustering network, is invested based on the kernelized semantic embedding learner. A triplet self-supervised optimization objective is designed in the DKEC model to guide the clustering partition, the separable factor of the kernelized semantic embedding, and the data reconstruction factor of the kernelized semantic embedding, simultaneously. We conducted extensive experiments on $ 7 $ real-world datasets. The results unequivocally showcase the exceptional clustering performance of the proposed DKEC. © 2024, The Authors. All rights reserved.

关键词： Deep neural networks

Burst-Sensitive Traffic Forecast Via Multi-Property Personalized Fusion in Federated Learning

学校读者我要写书评

暂无评论

IEEE Transactions on Mobile computing 2025年

作者： Xue, Jingjing Sun, Sheng Liu, Min Wang, Yuwei Meng, Xuying Wang, Jingyuan Zhang, JunBo Xu, Ke Chinese Academy of Sciences University of Chinese Academy of Sciences Institute of Computing Technology Beijing China Chinese Academy of Sciences Institute of Computing Technology Beijing China Beihang University School of Computer Science and Engineering China Beihang University MIIT Key Laboratory of Data Intelligence and Management Beijing China JD Technology JD iCity China JD Intelligent Cities Research China Tsinghua University Department of Computer Science and Technology Beijing China Zhongguancun Laboratory Beijing China

For distributed network traffic prediction with data localization and privacy protection, Federated Learning (FL) enables collaborative training without raw data exchange across Base Stations (BSs). Nevertheless, traffic across BSs exhibit inherently heterogeneous trend burst and smooth fluctuation properties, but existing FL methods model single-scale series from only one view, which cannot simultaneously capture diverse trend and fluctuation properties, especially distinct burst distributions. In this paper, we propose Personalized Federated Forecasting with Multi-property Self-fusion (P2FMS), which can represent multi-scale traffic properties from different views. With precise multi-property representations, a fusion-level prediction decision is learned for each client in a personalized manner to promptly sense traffic bursts and improve forecasting performance in non-IID settings. Specifically, P2FMS decomposes the traffic series into distinct time scales, based on which, we effectively extract closeness, period, and trend properties from different views. The closeness and period are embedded through global-view representations with spatial correlations, while non-stationary trends are individually fitted from the client-side view. Furthermore, a personalized combiner is designed to accurately quantify the proportion of general fluctuation raws (i.e., closeness and period) and specific trend property in predictions, which enables multi-property self-fusion for each client to accommodate heterogeneous traffic patterns and enhance prediction accuracy. Besides, an alternant training mechanism is introduced to optimize property representation and fusion modules with the convergence guarantee. Extensive experiments on real-world datasets show that P2FMS outperforms status quo methods in both prediction performance and convergence time. © 2002-2012 IEEE.

关键词： Weather forecasting

Unifying and Improving Graph Convolutional Neural Networks with Wavelet Denoising Filters 23

学校读者我要写书评

暂无评论

Unifying and Improving Graph Convolutional Neural Networks w...

2023 World Wide Web Conference, WWW 2023

作者： Wan, Liangtian Li, Xiaona Han, Huijin Yan, Xiaoran Sun, Lu Ning, Zhaolong Xia, Feng Key Laboratory for Ubiquitous Network and Service Software of Liaoning Province School of Software Dalian University of Technology Dalian China Research Center of Big Data Intelligence Research Institute of Artificial Intelligence Zhejiang Lab Hangzhou China Department of Communication Engineering Institute of Information Science Technology Dalian Maritime University Dalian China School of Communication and Information Engineering Chongqing University of Posts and Telecommunications Chongqing China School of Computing Technologies Rmit University Melbourne Australia

ISBN: (纸本)9781450394161

Graph convolutional neural network (GCN) is a powerful deep learning framework for network data. However, variants of graph neural architectures can lead to drastically different performance on different tasks. Model comparison calls for a unifying framework with interpretability and principled experimental procedures. Based on the theories from graph signal processing (GSP), we show that GCN's capability is fundamentally limited by the uncertainty principle, and wavelets provide a controllable trade-off between local and global information. We adapt wavelet denoising filters to the graph domain, unifying popular variants of GCN under a common interpretable mathematical framework. Furthermore, we propose WaveThresh and WaveShrink which are novel GCN models based on proven denoising filters from the signal processing literature. Empirically, we evaluate our models and other popular GCNs under a more principled procedure and analyze how trade-offs between local and global graph signals can lead to better performance in different datasets. © 2023 ACM.

关键词： Convolution

CIMI4D: A Large Multimodal Climbing Motion dataset under Human-scene Interactions

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Yan, Ming Wang, Xin Dai, Yudi Shen, Siqi Wen, Chenglu Xu, Lan Ma, Yuexin Wang, Cheng Fujian Key Laboratory of Sensing and Computing for Smart Cities Xiamen University China National Institute for Data Science in Health and Medicine Xiamen University China Key Laboratory of Multimedia Trusted Perception and Efficient Computing Ministry of Education of China School of Informatics Xiamen University China Shanghai Engineering Research Center of Intelligent Vision and Imaging ShanghaiTech University China

Motion capture is a long-standing research problem. Although it has been studied for decades, the majority of research focus on ground-based movements such as walking, sitting, dancing, etc. Off-grounded actions such as climbing are largely overlooked. As an important type of action in sports and firefighting field, the climbing movements is challenging to capture because of its complex back poses, intricate human-scene interactions, and difficult global localization. The research community does not have an in-depth understanding of the climbing action due to the lack of specific datasets. To address this limitation, we collect CIMI4D, a large rock ClImbing MotIon dataset from 12 persons climbing 13 different climbing walls. The dataset consists of around 180,000 frames of pose inertial measurements, LiDAR point clouds, RGB videos, high-precision static point cloud scenes, and reconstructed scene meshes. Moreover, we frame-wise annotate touch rock holds to facilitate a detailed exploration of human-scene interaction. The core of this dataset is a blending optimization process, which corrects for the pose as it drifts and is affected by the magnetic conditions. To evaluate the merit of CIMI4D, we perform four tasks which include human pose estimations (with/without scene constraints), pose prediction, and pose generation. The experimental results demonstrate that CIMI4D presents great challenges to existing methods and enables extensive research opportunities. We share the dataset with the research community in http://***/cimi4d/. Copyright © 2023, The Authors. All rights reserved.

关键词： Blending

A Fusion Tuning Method for Named Entity Recognition 12th

学校读者我要写书评

暂无评论

A Fusion Tuning Method for Named Entity Recognition

12th CCF Conference on Bigdata, Bigdata 2024

作者： Wang, Jitian Chen, Yanping Zou, Anqi Qin, Yongbin Huang, Ruizhang Text Computing and Cognitive Intelligence Engineering Research Center of National Education Ministry Guizhou University Guiyang550025 China State Key Laboratory of Public Big Data Guizhou University Guiyang China College of Computer Science and Technology Guizhou University Guiyang550025 China

ISBN: (纸本)9789819610235

In named entity recognition, the main methods for constructing deep neural networks are fine-tuning and prompt tuning. Fine-tuning is a commonly used paradigm to optimize neural networks by using task-specific objective functions to make them more adapted to solve downstream tasks. In which pre-trained language models are used as external resources to provide contextual representations. Prompt tuning is a recent paradigm in which the downstream tasks are reformulated to look more like those solved during the original PLM training with the help of a textual prompt. It is effective to utilize potential knowledge of PLMs. In this paper, we propose the fusion approach, which sufficiently fuses the semantics obtained by extracting fine-tuning and prompt-tuning. In our approach, the model makes decisions based on the information obtained from the reconstructive fusion of the fine-tuned semantics and the semantics of the pre-trained language model. The representation obtained by fine-tuning and the representation obtained by prompt tuning are reconstructed into a multidimensional semantics using a gate fusion module (GFM). In particularly, our model has two gate components, one gate to extract the semantics of fine-tuning and the other gate to extract the semantics of prompt tuning, and finally the reconstructed semantic representation is obtained. This takes full advantage of the potential knowledge of PLM. Our model is evaluated on the CONLL2003, ACE2005 and GENIA corpora. It achieves performance close to the state-of-the-art in F1 scores. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Semantics

NOBGP: A Novel Optimized Balanced Graph Partitioning Algorithm 19th

学校读者我要写书评

暂无评论

NOBGP: A Novel Optimized Balanced Graph Partitioning Algorit...

19th CCF Conference on Computer Supported Cooperative Work and Social computing, ChineseCSCW 2024

作者： Chen, Jiebin Hu, Ziqiang Ye, Renjie Zhang, Qishan Guo, Kun College of Computer and Data Science Fuzhou University Fuzhou350108 China Engineering Research Center of Big Data Intelligence Ministry of Education Fuzhou350108 China Fujian Key Laboratory of Network Computing and Intelligent Information Processing Fuzhou University Fuzhou350108 China Xianda College of Economics and Humanities Shanghai International Studies University Shanghai China

ISBN: (纸本)9789819623723

Large-scale graphs have become prevalent with the advent of the big data era. Distributed graph computing systems are commonly used for processing and analyzing large-scale graphs, with graph partitioning being a key prerequisite for their efficient computation. Graph partitioning aims to balance the load across partitions while minimizing the number of cut-edges. Moreover, it should achieve high efficiency and scalability. However, the existing popular graph partitioning algorithms do not fully take into account the internal topology of real-world graphs, which affects the final partition quality and convergence. Meanwhile, they easily fall into the local optimum due to partition load constraints. This paper introduces a Novel Optimized Balanced Graph Partitioning algorithm (NOBGP). First, we propose an initialization strategy based on label propagation of core vertices to achieve initial partitions with good locality and accelerate convergence. Second, we optimize the label propagation process to ensure balanced partitions and propose a probability-based disruption strategy to avoid the local optimum. We implement NOBGP on the distributed graph computing framework GraphX. Extensive experimental results on real-world graphs show that the proposed algorithm is scalable and performs better than the existing algorithms. We also run PageRank and Louvain applications using the graph partitioning results to demonstrate the efficiency of our algorithm. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Graph algorithms