检索结果-内蒙古大学图书馆

SSCNet:learning-based subspace clustering

Visual intelligence 2024年第1期2卷 116-131页

作者： Xingyu Xie Jianlong Wu Guangcan Liu Zhouchen Lin National Key Lab of General AI School of Intelligence Science and TechnologyPeking UniversityBeijingChina School of Computer Science and Technology Harbin Institute of Technology(Shenzhen)ShenzhenChina School of Automation Southeast UniversityNanjingChina Institute for Artificial Intelligence Peking UniversityBeijingChina Peng Cheng Laboratory ShenzhenChina

Sparse subspace clustering(SSC),a seminal clustering method,has demonstrated remarkable performance by effectively solving the data sparsity ***,it is not without its *** among these is the difficulty of incremental learning with the original SSC,accompanied by a computationally demanding recalculation process that constrains its scalability to large ***,the conventional SSC framework considers dictionary construction,affinity matrix learning and clustering as separate stages,potentially leading to suboptimal dictionaries and affinity matrices for *** address these challenges,we present a novel clustering approach,called SSCNet,which leverages differentiable ***,we redefine and generalize the optimization procedure of the linearized alternating direction method of multipliers(ADMM),framing it as a multi-block deep neural network,where each block corresponds to a linearized ADMM iteration *** reformulation is used to address the SSC *** then use a shallow spectral embedding network as an unambiguous and differentiable module to approximate the eigenvalue ***,we incorporate a self-supervised structure to mitigate the non-differentiability inherent in k-means to achieve the final clustering *** essence,we assign unique objectives to different modules and jointly optimize all module parameters using stochastic gradient *** to the high efficiency of the optimization process,SSCNet can be easily applied to large-scale *** evaluations on several benchmarks confirm that our method outperforms traditional state-of-the-art approaches.

关键词： Subspace clustering Learning-based optimization Linearized alternating direction method of multipliers(ADMM) Differentiable low-rank decomposition

来源：评论

学校读者我要写书评

暂无评论

Don’t Ignore the Drive of Curiosity: Rethinking Subtleties Between Universality of Commonsense Knowledge and Excellence of Large Language Models

引用

SN computer science 2024年第6期5卷 1-14页

作者： Wang, Chao Chen, Tao Liu, Jingping School of Future Technology Shanghai University Shanghai China Institute of Artificial Intelligence Shanghai University Shanghai China Shanghai Key Laboratory of Data Science School of Computer Science Fudan University Shanghai China School of Information Science and Engineering East China University of Science and Technology Shanghai China

Commonsense reasoning is one of the abilities necessary for artificial intelligence to be as intelligent as humans. However, how to make AI understand commonsense has been a problem that has plagued artificial intelligence for more than 60 years. Existing efforts focus more on the means of knowledge acquisition and strive to enrich the capacity of commonsense knowledge (CSK) bases and dimensions of CSK through advanced methods. Unfortunately, this exuberance has obscured a general consideration of CSK, such as how to follow human habits to obtain the most representative knowledge we need to understand the world. In this paper, this representative knowledge is referred to as core CSK. The influence of core CSK is extensive, and it constitutes almost the fundamental element of human life and the most fundamental cognition of the world. Harnessing human curiosity to find solutions to the above problems is an effective and straightforward route. Specifically, we focus on a special corpus to mine core CSK, namely, why-questions. For example, we can harvest “the sky is blue” from “why is the sky blue?”. To this end, we propose a novel method to extract CSK from why-questions, which mainly consist of two modules. The first is a question classification module used to determine whether a question contains CSK. In this module, we propose a classifier based on a one-sided bootstrapping method and design several informative features for the classifier. The second is a crowdsourcing module used to improve the quality of the extracted commonsense. We conduct extensive experiments, and the experimental results show that our method effectively mines CSK from question corpora. Furthermore, statistical analysis demonstrates the feasibility of this curiosity-driven approach, implying that we provide a basic idea for collecting core CSK. Remarkably, today’s outstanding large language models do not have such simple knowledge summarization capabilities, demonstrating the barrier between

关键词： Classification Core Commonsense knowledge Crowdsourcing Curiosity Psychologically

来源：评论

学校读者我要写书评

暂无评论

A detection method for image content aware tampering based on ULBP

A detection method for image content aware tampering based o...

引用

2023 International Conference on computer Application and Information Security, ICCAIS 2023

作者： Lin, Yiran Lu, Ming Key Laboratory of Intelligent Informatics for Safety & Emergency of Zhejiang Province Wenzhou University Zhejiang Wenzhou325035 China College of Computer Science and Artificial Intelligence Wenzhou University Zhejiang Wenzhou325035 China

ISBN: (纸本)9781510675032

A digital image tampering detection method based on a combination of unified local binary mode features and energy deviation features is proposed for digital image tampering achieved using content aware scaling technology. This method describes the changes in pixel correlation within local neighborhoods caused by image tampering by unifying local binary mode features, and achieves the same detection effect as traditional local binary modes using fewer dimensional features. Considering the changes in energy deviation of the image before and after tampering, a unified local binary pattern feature and energy deviation feature are combined to train the classifier using joint features, thereby achieving more accurate and efficient detection results. The experimental results show that this method can effectively detect content aware tampering in images and accurately locate tampered areas in Seam Insertion. © 2024 SPIE.

关键词： Local binary pattern

来源：评论

学校读者我要写书评

暂无评论

Multilingual Brain Surgeon: Large Language Models Can be Compressed Leaving No Language Behind 30

Multilingual Brain Surgeon: Large Language Models Can be Com...

引用

Joint 30th International Conference on Computational Linguistics and 14th International Conference on Language Resources and Evaluation, LREC-COLING 2024

作者： Zeng, Hongchuan Xu, Hongshen Chen, Lu Yu, Kai X-LANCE Lab Department of Computer Science and Engineering MoE Key Lab of Artificial Intelligence SJTU AI Institute Shanghai Jiao Tong University Shanghai China Suzhou Laboratory Suzhou China

ISBN: (纸本)9782493814104

Large Language Models (LLMs) have ushered in a new era in Natural Language Processing, but their massive size demands effective compression techniques for practicality. Although numerous model compression techniques have been investigated, they typically rely on a calibration set that overlooks the multilingual context and results in significant accuracy degradation for low-resource languages. This paper introduces Multilingual Brain Surgeon (MBS), a novel calibration data sampling method for multilingual LLMs compression. MBS overcomes the English-centric limitations of existing methods by sampling calibration data from various languages proportionally to the language distribution of the model training datasets. Our experiments, conducted on the BLOOM multilingual LLM, demonstrate that MBS improves the performance of existing English-centric compression methods, especially for low-resource languages. We also uncover the dynamics of language interaction during compression, revealing that the larger the proportion of a language in the training set and the more similar the language is to the calibration language, the better performance the language retains after compression. In conclusion, MBS presents an innovative approach to compressing multilingual LLMs, addressing the performance disparities and improving the language inclusivity of existing compression techniques. The codes are available at: https://***/X-LANCE/MBS. © 2024 ELRA Language Resource Association: CC BY-NC 4.0.

关键词： Calibration

来源：评论

学校读者我要写书评

暂无评论

Rethinking the Membrane Dynamics and Optimization Objectives of Spiking Neural Networks 38

Rethinking the Membrane Dynamics and Optimization Objectives...

引用

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Shen, Hangchi Zheng, Qian Wang, Huamin Pan, Gang College of Artificial Intelligence Southwest University China Chongqing Key Laboratory of Brain Inspired Computing and Intelligent Chips China The State Key Lab of Brain-Machine Intelligence Zhejiang University China College of Computer Science and Technology Zhejiang University China

Despite spiking neural networks (SNNs) have demonstrated notable energy efficiency across various fields, the limited firing patterns of spiking neurons within fixed time steps restrict the expression of information, which impedes further improvement of SNN performance. In addition, current implementations of SNNs typically consider the firing rate or average membrane potential of the last layer as the output, lacking exploration of other possibilities. In this paper, we identify that the limited spike patterns of spiking neurons stem from the initial membrane potential (IMP), which is set to 0. By adjusting the IMP, the spiking neurons can generate additional firing patterns and pattern mappings. Furthermore, we find that in static tasks, the accuracy of SNNs at each time step increases as the membrane potential evolves from zero. This observation inspires us to propose a learnable IMP, which can accelerate the evolution of membrane potential and enables higher performance within a limited number of time steps. Additionally, we introduce the last time step (LTS) approach to accelerate convergence in static tasks, and we propose a label smooth temporal efficient training (TET) loss to mitigate the conflicts between optimization objective and regularization term in the vanilla TET. Our methods improve the accuracy by 4.05% on ImageNet compared to baseline and achieve state-of-the-art performance of 87.80% on CIFAR10-DVS and 87.86% on N-Caltech101. The code is available at https://***/StephenTaylor1998/IMP-SNN. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

CalPal: An Intelligent Multimodal Digital Wall Calendar

CalPal: An Intelligent Multimodal Digital Wall Calendar

引用

Joint of the ACM Workshops at the International Conference on Intelligent User Interfaces 2025, IUI-WS 2025

作者： Murphy, Devin Moralejo, Jenny Liang, Paul Pu Computer Science and Artificial Intelligence Laboratory Massachusetts Institute of Technology United States Massachusetts Institute of Technology United States Media Lab and EECS Massachusetts Institute of Technology United States

Mobile calendaring applications have become indispensable tools for organizing daily life, yet they lack the tactile and personalized qualities of traditional paper calendars, which have been shown to enhance engagement and goal commitment. While digital wall planners exist, they fail to replicate the rich, interactive experience of their physical counterparts. Leveraging recent advancements in generative artificial intelligence, we present CalPal, an intelligent multimodal digital wall calendar designed to merge the strengths of digital and paper-based planning. CalPal supports pen-based input for event creation in the user's own handwriting, gesture-based navigation for seamless interaction, and synchronization with mobile platforms. Additionally, it uses a stable diffusion model to create dynamic, visually engaging themes based on the calendar's events, aiming to enhance the reflective and aesthetic qualities of the calendaring experience. We outline the design and implementation of CalPal, illustrate its functionality through an example user scenario, and discuss future work to evaluate the effectiveness of the platform in harmonizing the strengths of traditional and digital calendaring practices. © 2025 Copyright © 2025 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0)

关键词： Character recognition

来源：评论

学校读者我要写书评

暂无评论

Dynamic Individual Selection and Crossover Boosted Forensic-based Investigation Algorithm for Global Optimization and Feature Selection

引用

Journal of Bionic Engineering 2023年第5期20卷 2416-2442页

作者： Hanyu Hu Weifeng Shan Jun Chen Lili Xing Ali Asghar Heidari Huiling Chen Xinxin He Maofa Wang School of Emergency Management Institute of Disaster PreventionSanhe065201China Earthquake Administration of Anhui Province Hefei230031China Department of Computer Science and Artificial Intelligence Wenzhou UniversityWenzhou325035China Guangxi Key Laboratory of Trusted Software Guilin University of Electronic TechnologyGuilin541004China

The advent of Big Data has rendered Machine Learning tasks more intricate as they frequently involve higher-dimensional *** Selection(FS)methods can abate the complexity of the data and enhance the accuracy,generalizability,and interpretability of ***-heuristic algorithms are often utilized for FS tasks due to their low requirements and efficient *** paper introduces an augmented Forensic-Based Investigation algorithm(DCFBI)that incorporates a Dynamic Individual Selection(DIS)and crisscross(CC)mechanism to improve the pursuit phase of the ***,a binary version of DCFBI(BDCFBI)is applied to *** conducted on IEEE CEC 2017 with other metaheuristics demonstrate that DCFBI surpasses them in search *** influence of different mechanisms on the original FBI is analyzed on benchmark functions,while its scalability is verified by comparing it with the original FBI on benchmarks with varied *** is then applied to 18 real datasets from the UCI machine learning database and the Wieslaw dataset to select near-optimal features,which are then compared with six renowned binary *** results show that BDCFBI can be more competitive than similar methods and acquire a subset of features with superior classification accuracy.

关键词： Feature selection Forensic-based investigation algorithm Crisscross mechanism Global optimization Metaheuristic algorithms Bionic algorithm

来源：评论

学校读者我要写书评

暂无评论

Learning to Communicate Among Agents for Large-Scale Dynamic Path Planning With Genetic Programming Hyperheuristic

IEEE Transactions on Artificial Intelligence

引用

IEEE Transactions on artificial intelligence 2025年第5期6卷 1269-1283页

作者： Liao, Xiao-Cheng Hu, Xiao-Min Chen, Xiang-Ling Mei, Yi Jia, Ya-Hui Chen, Wei-Neng Victoria University of Wellington Centre for Data Science and Artificial Intelligence School of Engineering and Computer Science Wellington6140 New Zealand Guangdong University of Technology School of Computer Science and Technology Guangzhou510006 China Hanyang University Department of Electrical and Electronic Engineering Ansan15588 Korea Republic of South China University of Technology School of Future Technology Guangzhou510006 China Pazhou Lab Guangzhou510005 China South China University of Technology School of Computer Science and Engineering State Key Laboratory of Subtropical Building and Urban Science Guangzhou510006 China

Genetic programming hyperheuristic (GPHH) has recently become a promising methodology for large-scale dynamic path planning (LDPP) since it can produce reusable heuristics rather than disposable solutions. However, in this methodology, the extracted local and decentralized heuristic for agents that lack a global systemic view sometimes may be problematic. Therefore, a new challenge is to strike a balance between conciseness to improve generalization ability and incorporation of more global information to obtain better performance. In this work, we target the LDPP problem and propose a communication learning mechanism (ComLGP) for GPHH to address the above difficulties. In ComLGP, a communication function is introduced to serve as a communication protocol and exist in the form of an extra terminal in GPHH. Compared to the classic terminals which are fixed in genetic programing, this communication function undergoes optimization along with the evolutionary process of GPHH. In this way, the communication function can be learned which enables agents to communicate without a predefined communication protocol. Then, a caching and lazy updating mechanism for ComLGP is presented to accelerate the calculation of communication content. Last, we verified our method on 22 scenarios including two real world road networks. The experimental results demonstrate that the proposed ComLGP can successfully learn to communicate. Although in the absence of any manually designed communication features, ComLGP is capable of achieving performance competitive to the state-of-the-art method that employs a predefined communication protocol and outperforms the remaining compared methods in most scenarios. © 2020 IEEE.

关键词： Genetic programming

来源：评论

学校读者我要写书评

暂无评论

3D Reconstruction and Texture Optimization Based on Differentiable Rendering 8

3D Reconstruction and Texture Optimization Based on Differen...

引用

8th International Conference on Intelligent Computing and Signal Processing, ICSP 2023

作者： Liu, Yongxin Zhao, Haifeng Zhang, Shaojie Fu, Yanping Anhui University Anhui Provincial Key Laboratory of Multimodal Cognitive Computation School of Computer Science and Technology Institute of Artificial Intelligence Hefei Comprehensive National Science Center Hefei China

ISBN: (纸本)9798350302455

Due to noise and quantization errors inherent in depth cameras, RGB-D based 3D reconstruction algorithms inevitably suffer from defects such as texture blurring and ghosting. However, current texture optimization methods tend to only focus on optimizing one aspect, such as camera pose, geometry, or texture, without considering the highly coupled relationship between them. As a result, it is difficult to achieve global optimality. To address this issue, we propose a joint optimization framework based on differentiable rendering that simultaneously optimizes camera pose, geometry, and texture. By fully exploiting the complementary relationship between these aspects, we are able to reconstruct textures with high fidelity while recovering fine geometric details. First, we propose an optimized geometry model combined with adaptive subdivision, where we selectively subdivide the model only in texture-rich regions to increase the number of vertex facets. Then, we perform geometry optimization to recover high-frequency geometric details evenly in the presence of large geometric errors without incurring computational costs. After the pose correction and geometric model optimization, we use pixel regeneration pipelines to generate clear texture atlases. Comparative evaluations with state-of-the-art texture optimization methods on a public dataset demonstrate the effectiveness of our proposed approach in reducing errors in the reconstruction of 3D models, generating clear textures, and recovering high-frequency geometric details. © 2023 IEEE.

关键词： Textures

来源：评论

学校读者我要写书评

暂无评论

Generated and Pseudo Content guided Prototype Refinement for Few-shot Point Cloud Segmentation 38

Generated and Pseudo Content guided Prototype Refinement for...

引用

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Wei, Lili Lang, Congyan Chen, Ziyi Wang, Tao Li, Yidong Liu, Jun School of Computer Science & Technology Beijing Jiaotong University China Key Laboratory of Big Data & Artificial Intelligence in Transportation Ministry of Education China School of Computing and Communications Lancaster University United Kingdom

Few-shot 3D point cloud semantic segmentation aims to segment query point clouds with only a few annotated support point clouds. Existing prototype-based methods learn prototypes from the 3D support set to guide the segmentation of query point clouds. However, they encounter the challenge of low prototype quality due to constrained semantic information in the 3D support set and class information bias between support and query sets. To address these issues, in this paper, we propose a novel framework called Generated and Pseudo Content guided Prototype Refinement (GPCPR), which explicitly leverages LLM-generated content and reliable query context to enhance prototype quality. GPCPR achieves prototype refinement through two core components: LLM-driven Generated Content-guided Prototype Refinement (GCPR) and Pseudo Query Context-guided Prototype Refinement (PCPR). Specifically, GCPR integrates diverse and differentiated class descriptions generated by large language models to enrich prototypes with comprehensive semantic knowledge. PCPR further aggregates reliable class-specific pseudo-query context to mitigate class information bias and generate more suitable query-specific prototypes. Furthermore, we introduce a dual-distillation regularization term, enabling knowledge transfer between early-stage entities (prototypes or pseudo predictions) and their deeper counterparts to enhance refinement. Extensive experiments demonstrate the superiority of our method, surpassing the state-of-the-art methods by up to 12.10% and 13.75% mIoU on S3DIS and ScanNet, respectively. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：