检索结果-内蒙古大学图书馆

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

DEGSTalk: Decomposed Per-Embedding Gaussian Fields for Hair-Preserving Talking Face Synthesis

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Deng, Kaijun Zheng, Dezhi Xie, Jindong Wang, Jinbao Xie, Weicheng Shen, Linlin Song, Siyang Computer Vision Institute School of Computer Science and Software Engineering Shenzhen University China National Engineering Laboratory for Big Data System Computing Technology Shenzhen University China Guangdong Provincial Key Laboratory of Intelligent Information Processing China Department of Computer Science University of Exeter United Kingdom

Accurately synthesizing talking face videos and capturing fine facial features for individuals with long hair presents a significant challenge. To tackle these challenges in existing methods, we propose a decomposed per-embedding Gaussian fields (DEGSTalk), a 3D Gaussian Splatting (3DGS)-based talking face synthesis method for generating realistic talking faces with long hairs. Our DEGSTalk employs Deformable Pre-Embedding Gaussian Fields, which dynamically adjust pre-embedding Gaussian primitives using implicit expression coefficients. This enables precise capture of dynamic facial regions and subtle expressions. Additionally, we propose a Dynamic Hair-Preserving Portrait Rendering technique to enhance the realism of long hair motions in the synthesized videos. Results show that DEGSTalk achieves improved realism and synthesis quality compared to existing approaches, particularly in handling complex facial dynamics and hair preservation. Our code will be publicly available at https://***/CVI-SZU/DEGSTalk. Copyright © 2024, The Authors. All rights reserved.

关键词： Gaussian distribution

HairDiffusion: Vivid Multi-Colored Hair Editing via Latent Diffusion

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Zeng, Yu Zhang, Yang Liu, Jiachen Shen, Linlin Deng, Kaijun He, Weizhao Wang, Jinbao Computer Vision Institute School of Computer Science & Software Engineering Shenzhen University China Shenzhen Institute of Artificial Intelligence and Robotics for Society China National Engineering Laboratory for Big Data System Computing Technology Shenzhen University China Guangdong Provincial Key Laboratory of Intelligent Information Processing China

Hair editing is a critical image synthesis task that aims to edit hair color and hairstyle using text descriptions or reference images, while preserving irrelevant attributes (e.g., identity, background, cloth). Many existing methods are based on StyleGAN to address this task. However, due to the limited spatial distribution of StyleGAN, it struggles with multiple hair color editing and facial preservation. Considering the advancements in diffusion models, we utilize Latent Diffusion Models (LDMs) for hairstyle editing. Our approach introduces Multi-stage Hairstyle Blend (MHB), effectively separating control of hair color and hairstyle in diffusion latent space. Additionally, we train a warping module to align the hair color with the target region. To further enhance multi-color hairstyle editing, we fine-tuned a CLIP model using a multi-color hairstyle dataset. Our method not only tackles the complexity of multi-color hairstyles but also addresses the challenge of preserving original colors during diffusion editing. Extensive experiments showcase the superiority of our method in editing multi-color hairstyles while preserving facial attributes given textual descriptions and reference images. © 2024, CC BY.

关键词： Color image processing

Detecting Adversarial data by Probing Multiple Perturbations Using Expected Perturbation Score

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Zhang, Shuhai Liu, Feng Yang, Jiahao Yang, Yifan Li, Changsheng Han, Bo Tan, Mingkui School of Software Engineering South China University of Technology China Pazhou Laboratory China The University of Melbourne Australia Beijing Institute of Technology Beijing China Department of Computer Science Hong Kong Baptist University Hong Kong Key Laboratory of Big Data and Intelligent Robot Ministry of Education China

Adversarial detection aims to determine whether a given sample is an adversarial one based on the discrepancy between natural and adversarial distributions. Unfortunately, estimating or comparing two data distributions is extremely difficult, especially in high-dimension spaces. Recently, the gradient of log probability density (a.k.a., score) w.r.t. the sample is used as an alternative statistic to compute. However, we find that the score is sensitive in identifying adversarial samples due to insufficient information with one sample only. In this paper, we propose a new statistic called expected perturbation score (EPS), which is essentially the expected score of a sample after various perturbations. Specifically, to obtain adequate information regarding one sample, we perturb it by adding various noises to capture its multi-view observations. We theoretically prove that EPS is a proper statistic to compute the discrepancy between two samples under mild conditions. In practice, we can use a pre-trained diffusion model to estimate EPS for each sample. Last, we propose an EPS-based adversarial detection (EPSAD) method, in which we develop EPS-based maximum mean discrepancy (MMD) as a metric to measure the discrepancy between the test sample and natural samples. We also prove that the EPS-based MMD between natural and adversarial samples is larger than that among natural samples. Extensive experiments show the superior adversarial detection performance of our EPS-AD. Copyright © 2023, The Authors. All rights reserved.

关键词： Machine learning

Source Code Summarization in the Era of Large Language Models

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Sun, Weisong Miao, Yun Li, Yuekang Zhang, Hongyu Fang, Chunrong Liu, Yi Deng, Gelei Liu, Yang Chen, Zhenyu State Key Laboratory for Novel Software Technology Nanjing University Nanjing China College of Computing and Data Science Nanyang Technological University Singapore Singapore School of Computer Science and Engineering University of New South Wales Sidney Australia School of Big Data and Software Engineering Chongqing University Chongqing China

To support software developers in understanding and maintaining programs, various automatic (source) code summarization techniques have been proposed to generate a concise natural language summary (i.e., comment) for a given code snippet. Recently, the emergence of large language models (LLMs) has led to a great boost in the performance of code-related tasks. In this paper, we undertake a systematic and comprehensive study on code summarization in the era of LLMs, which covers multiple aspects involved in the workflow of LLM-based code summarization. Specifically, we begin by examining prevalent automated evaluation methods for assessing the quality of summaries generated by LLMs and find that the results of the GPT-4 evaluation method are most closely aligned with human evaluation. Then, we explore the effectiveness of five prompting techniques (zero-shot, few-shot, chain-of-thought, critique, and expert) in adapting LLMs to code summarization tasks. Contrary to expectations, advanced prompting techniques may not outperform simple zero-shot prompting. Next, we investigate the impact of LLMs’ model settings (including top_p and temperature parameters) on the quality of generated summaries. We find the impact of the two parameters on summary quality varies by the base LLM and programming language, but their impacts are similar. Moreover, we canvass LLMs’ abilities to summarize code snippets in distinct types of programming languages. The results reveal that LLMs perform suboptimally when summarizing code written in logic programming languages compared to other language types (e.g., procedural and object-oriented programming languages). Finally, we unexpectedly find that CodeLlama-Instruct with 7B parameters can outperform advanced GPT-4 in generating summaries describing code implementation details and asserting code properties. We hope that our findings can provide a comprehensive understanding of code summarization in the era of *** Codes 68-04 Copyright © 202

关键词： Quality control

ALFLAT: Chinese NER Using ALBERT, Flat-Lattice Transformer, Word Segmentation and Entity Dictionary 2nd

学校读者我要写书评

暂无评论

ALFLAT: Chinese NER Using ALBERT, Flat-Lattice Transformer, ...

2nd EAI International Conference on Applied Cryptography in Computer and Communications, AC3 2022

作者： Lv, Haifeng Ding, Yong School of Data Science and Software Engineering Wuzhou University Wuzhou China Guangxi Key Laboratory of Cryptography and Information Security School of Computer Science and Information Security Guilin University of Electronic Technology Guilin China Guangxi Key Laboratory of Machine Vision and Intelligent Control Wuzhou University Wuzhou China

ISBN: (纸本)9783031170805

Recently, the character-word lattice structure has been proved to be effective for Chinese named entity recognition (NER) by incorporating the word information. However, one hand, since the lattice structure is dynamic and complex, although some existing lattice-based models are effectively utilize the parallel computation of GPUs, they do not fully utilize word segmentation boundary tags that as features are helpful for NER task. On the other hand, the character-word vector needs to be trained, and the user-defined entity dictionary cannot be effectively used. In this paper, we propose ALFLAT: based on a flat-lattice Transformer to incorporate ALBERT pre-trained model, word segmentation information and user-defined entity dictionary for Chinese NER. ALFLAT converts the lattice structure into a flat structure consisting of spans, integrate word segmentation embedding with the output of flat-lattice Transformer model, then modifies the emission scores according to the user-defined entity dictionary, finally utilize Viterbi decoding of the CRF layer to obtain the correct entity results. Each span corresponds to a character or latent word and its position in the original lattice. With the power of ALBERT pre-trained model, Transformer and position encoding, ALFLAT can fully leverage the lattice, word segmentation and user-defined entity dictionary information. Experiments on MSRA dataset show ALFLAT outperforms other lexicon-based models in performance and efficiency. © 2022, ICST Institute for Computer Sciences, Social Informatics and Telecommunications engineering.

关键词： Computational linguistics

Diffusion-Based mmWave Radar Point Cloud Enhancement Driven by Range Images

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Wu, Ruixin Li, Zihan Wang, Jin Xu, Xiangyu Yu, Huan Zheng, Zhi Huang, Kaixiang Lu, Guodong The State Key Laboratory of Fluid Power and Mechatronic Systems School of Mechanical Engineering Zhejiang University Hangzhou310058 China Zhejiang Key Laboratory of Industrial Big Data and Robot Intelligent Systems Zhejiang University Hangzhou310058 China Robotics Research Center of Yuyao City Ningbo315400 China State Key Laboratory of Robotics and Systems Department of Mechatronics Engineering Harbin Institute of Technology Harbin150001 China

Millimeter-wave (mmWave) radar has attracted significant attention in robotics and autonomous driving. However, despite the perception stability in harsh environments, the point cloud generated by mmWave radar is relatively sparse while containing significant noise, which limits its further development. Traditional mmWave radar enhancement approaches often struggle to leverage the effectiveness of diffusion models in super-resolution, largely due to the unnatural range-azimuth heatmap (RAH) or bird’s eye view (BEV) representation. To overcome this limitation, we propose a novel method that pioneers the application of fusing range images with image diffusion models, achieving accurate and dense mmWave radar point clouds that are similar to LiDAR. Benefitting from the projection that aligns with human observation, the range image representation of mmWave radar is close to natural images, allowing the knowledge from pre-trained image diffusion models to be effectively transferred, significantly improving the overall performance. Extensive evaluations on both public datasets and self-constructed datasets demonstrate that our approach provides substantial improvements, establishing a new state-of-the-art performance in generating truly three-dimensional LiDAR-like point clouds via mmWave radar. Copyright © 2025, The Authors. All rights reserved.

关键词： Radar imaging

Efficient Test-Time Model Adaptation without Forgetting

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Niu, Shuaicheng Wu, Jiaxiang Zhang, Yifan Chen, Yaofo Zheng, Shijian Zhao, Peilin Tan, Mingkui School of Software Engineering South China University of Technology China Key Laboratory of Big Data and Intelligent Robot Ministry of Education China Tencent AI Lab China National University of Singapore Singapore Pazhou Laboratory China

Test-time adaptation (TTA) seeks to tackle potential distribution shifts between training and testing data by adapting a given model w.r.t. any testing sample. This task is particularly important for deep models when the test environment changes frequently. Although some recent attempts have been made to handle this task, we still face two practical challenges: 1) existing methods have to perform backward computation for each test sample, resulting in unbearable prediction cost to many applications;2) while existing TTA solutions can significantly improve the test performance on out-of-distribution data, they often suffer from severe performance degradation on in-distribution data after TTA (known as catastrophic forgetting). In this paper, we point out that not all the test samples contribute equally to model adaptation, and high-entropy ones may lead to noisy gradients that could disrupt the model. Motivated by this, we propose an active sample selection criterion to identify reliable and non-redundant samples, on which the model is updated to minimize the entropy loss for test-time adaptation. Furthermore, to alleviate the forgetting issue, we introduce a Fisher regularizer to constrain important model parameters from drastic changes, where the Fisher importance is estimated from test samples with generated pseudo labels. Extensive experiments on CIFAR-10-C, ImageNet-C, and ImageNet-R verify the effectiveness of our proposed method. Code is available at https://***/mr-eggplant/EATA. Copyright © 2022, The Authors. All rights reserved.

关键词： Entropy

Confidence-aware Contrastive Learning for Selective Classification

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Wu, Yu-Chang Lyu, Shen-Huan Shang, Haopu Wang, Xiangyu Qian, Chao National Key Laboratory for Novel Software Technology Nanjing University China School of Artificial Intelligence Nanjing University China Key Laboratory of Water Big Data Technology of Ministry of Water Resources Hohai University China College of Computer Science and Software Engineering Hohai University China

Selective classification enables models to make predictions only when they are sufficiently confident, aiming to enhance safety and reliability, which is important in high-stakes scenarios. Previous methods mainly use deep neural networks and focus on modifying the architecture of classification layers to enable the model to estimate the confidence of its prediction. This work provides a generalization bound for selective classification, disclosing that optimizing feature layers helps improve the performance of selective classification. Inspired by this theory, we propose to explicitly improve the selective classification model at the feature level for the first time, leading to a novel Confidence-aware Contrastive Learning method for Selective Classification, CCL-SC, which similarizes the features of homogeneous instances and differentiates the features of heterogeneous instances, with the strength controlled by the model’s confidence. The experimental results on typical datasets, i.e., CIFAR-10, CIFAR-100, CelebA, and ImageNet, show that CCL-SC achieves significantly lower selective risk than state-of-the-art methods, across almost all coverage degrees. Moreover, it can be combined with existing methods to bring further improvement. © 2024, CC BY.

关键词： Classification (of information)