检索结果-内蒙古大学图书馆

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Shi, Luohe Yao, Yao Li, Zuchao Zhang, Lefei Zhao, Hai National Engineering Research Center for Multimedia Software School of Computer Science Wuhan University Wuhan430072 China Department of Computer Science and Engineering Shanghai Jiao Tong University China

Large language models (LLMs) have rapidly advanced and demonstrated impressive capabilities. In-Context Learning (ICL) and Parameter-Efficient Fine-Tuning (PEFT) are currently two mainstream methods for augmenting LLMs to downstream tasks. ICL typically constructs a few-shot learning scenario, either manually or by setting up a Retrieval-Augmented Generation (RAG) system, helping models quickly grasp domain knowledge or question-answering patterns without changing model parameters. However, this approach involves trade-offs, such as slower inference speed and increased space occupancy. PEFT assists the model in adapting to tasks through minimal parameter modifications, but the training process still demands high hardware requirements, even with a small number of parameters involved. To address these challenges, we propose Reference Trustable Decoding (RTD), a paradigm that allows models to quickly adapt to new tasks without fine-tuning, maintaining low inference costs. RTD constructs a reference datastore from the provided training examples and optimizes the LLM's final vocabulary distribution by flexibly selecting suitable references based on the input, resulting in more trustable responses and enabling the model to adapt to downstream tasks at a low cost. Experimental evaluations on various LLMs using different benchmarks demonstrate that RTD establishes a new paradigm for augmenting models to downstream tasks. Furthermore, our method exhibits strong orthogonality with traditional methods, allowing for concurrent usage. Our code can be found at https://***/ShiLuohe/ReferenceTrustableDecoding. © 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

An improved pulse coupled neural networks model for semantic IoT

引用

Digital Communications and Networks 2024年第3期10卷 557-567页

作者： Rong Ma Zhen Zhang Yide Ma Xiping Hu Edith C.H.Ngai Victor C.M.Leung School of Information Science and Engineering Lanzhou UniversityLanzhou 730000China School of Medical Technology Beijing Institute of TechnologyBeijing100081China Department of Electrical and Electronic Engineer The University of Hong KongHong Kong 999077China College of Computer Science and Software Engineering Shenzhen UniversityShenzhen 518060China Department of Electrical and Computer Engineering the University of British ColumbiaVancouverBC V6T 1Z4Canada

In recent years,the Internet of Things(IoT)has gradually developed applications such as collecting sensory data and building intelligent services,which has led to an explosion in mobile data ***,with the rapid development of artificial intelligence,semantic communication has attracted great attention as a new communication ***,for IoT devices,however,processing image information efficiently in real time is an essential task for the rapid transmission of semantic *** the increase of model parameters in deep learning methods,the model inference time in sensor devices continues to *** contrast,the Pulse Coupled Neural Network(PCNN)has fewer parameters,making it more suitable for processing real-time scene tasks such as image segmentation,which lays the foundation for real-time,effective,and accurate image ***,the parameters of PCNN are determined by trial and error,which limits its *** overcome this limitation,an Improved Pulse Coupled Neural Networks(IPCNN)model is proposed in this *** IPCNN constructs the connection between the static properties of the input image and the dynamic properties of the neurons,and all its parameters are set adaptively,which avoids the inconvenience of manual setting in traditional methods and improves the adaptability of parameters to different types of *** segmentation results demonstrate the validity and efficiency of the proposed self-adaptive parameter setting method of IPCNN on the gray images and natural images from the Matlab and Berkeley Segmentation *** IPCNN method achieves a better segmentation result without training,providing a new solution for the real-time transmission of image semantic information.

关键词： Internet of things(IoT) Semantic information Real-time application Improved pulse coupled neural network Image segmentation

来源：评论

学校读者我要写书评

暂无评论

Two-Stage Lesion Detection Approach Based on Dimension-Decomposition and 3D Context

引用

Tsinghua science and Technology 2022年第1期27卷 103-113页

作者： Jiacheng Jiao Haiwei Pan Chunling Chen Tao Jin Yang Dong Jingyi Chen the College of Computer Science and Technology Harbin Engineering UniversityHarbin 150001China the School of Software Tsinghua UniversityBeijng 100084China

Lesion detection in Computed Tomography(CT) images is a challenging task in the field of computer-aided *** important issue is to locate the area of lesion *** a branch of Convolutional Neural Networks(CNNs),3D Context-Enhanced(3DCE) frameworks are designed to detect lesions on CT *** False Positives(FPs) detected in 3DCE frameworks are usually caused by inaccurate region proposals,which slow down the inference *** solve the above problems,a new method is proposed,a dimension-decomposition region proposal network is integrated into 3DCE framework to improve the location accuracy in lesion *** the restriction of "anchors" on ratios and scales,anchors are decomposed to independent "anchor strings".Anchor segments are dynamically combined in accordance with probability,and anchor strings with different lengths dynamically compose bounding *** show that the accurate region proposals generated by our model promote the sensitivity of FPs and spend less inference time compared with the current methods.

关键词： lesion detection Computed Tomography(CT) dimension-decomposition 3D context computer-aided diagnosis

来源：评论

学校读者我要写书评

暂无评论

Distributed Hierarchical Multi-Agent Reinforcement Learning for Real-World Adversarial Team Games 21

Distributed Hierarchical Multi-Agent Reinforcement Learning ...

引用

21st International computer Conference on Wavelet Active Media Technology and Information Processing, ICCWAMTIP 2024

作者： Wei, Li Bei, Hui Ke, Lu School of Computer Science and Engineering University of Electronic Science and Technology of China Chengdu611731 China School of Information and Software Engineering University of Electronic Science and Technology of China Chengdu611731 China

ISBN: (纸本)9798331519254

Real-world applications often require teams of agents to coordinate strategies against adversaries. However, existing MARL models overlook resource constraints and task complexity, limiting their practical use. To address this, we propose the Distributed Hierarchical Multi-agent Learning (DHML) framework, which uses distributed neural networks for efficient computation. A mean-field mechanism and an attentional communication module are further incorporated for scalability, and a hierarchical task decomposition is adopted based on the option-critic framework for handling complexity. Experiments on a simulation platform modeling real communication networks demonstrate DHML's effectiveness across various adversarial scenarios. © 2024 IEEE.

关键词： Adversarial machine learning

来源：评论

学校读者我要写书评

暂无评论

An Optimization Approach for Automata Strategies in Industrial Simulation 24

An Optimization Approach for Automata Strategies in Industri...

引用

8th International Conference on Control engineering and Artificial Intelligence, CCEAI 2024

作者： Zhang, Yuhan Chen, Zhendong Pan, Maolin Yu, Yang School of Computer Science and Engineering Sun Yat-Sen University Guangzhou China School of Software Engineering Sun Yat-Sen University Zhuhai China

ISBN: (纸本)9798400707971

In industrial contexts, managing time and resources typically involves complex systems with various operational components. This paper introduces a state machine model, optimized for simulating multi-component collaborative processes in industry, which is based on the State Chart XML (SCXML) standard and referred to as SCXML automata. The execution of these automata is treated as Markov decision processes, providing a framework for optimization. The primary goal is to develop strategies ensuring safe operation of such automata while adhering to time and cost constraints. A detailed representation of these automata is presented, including grammar, semantics, and strategies, and methods for strategy generation are adapted from single to multi-component systems. The approach involves the analysis of simulation data from such automata using adaptive reinforcement learning techniques, leading to the generation of strategies that approach optimality. Results demonstrate the capability of this approach in finding near-optimal strategies within a specified number of iterations. This method holds significant research value in the field of industrial flexible production lines, offering a means to generate strategies for multi-layered automata, thereby enhancing process efficiency. © 2024 ACM.

关键词： Optimization

来源：评论

学校读者我要写书评

暂无评论

A Deep Understanding Video Q&A System for Film Education in Acting Department

A Deep Understanding Video Q&A System for Film Education in ...

引用

2023 International Conference on Intelligent Education and Intelligent Research, IEIR 2023

作者： Wu, Zhengqian Li, Ruizhe Guo, Jiahao Wang, Zhongyuan Liang, Chao Hubei Key Laboratory of Multimedia and Network Communication Engineering School of Computer Science China National Engineering Research Center for Multimedia Software School of Computer Science China Wuhan University School of Computer Science China

ISBN: (纸本)9798350342895

Recently, advancements in artificial intelligence technology have greatly influenced the field of education, particularly in the area of intelligent homework assistance. However, current approaches are primarily designed for procedural and logical tasks and often lack comprehension abilities. This limitation is particularly evident when it comes to multi-hop and continuous tasks. To address this challenge, the integration of Large Language Model (LLM) has significantly enhanced the capability of AI systems to handle multi-hop and highly interconnected inputs. In this study, we focus on the learning needs of students in Acting Department, specifically their study of movies and the significance of classic movie videos in their learning process. However, assessing deep comprehension of classic movies poses its own challenges. To overcome these challenges, we develop a quiz system utilizing Knowledge Graphs (KG) and LLM to facilitate a deeper understanding of classic films. The generation of video quiz pairs is achieved through the use of Automatic Speech Recognition (ASR) technology, which leverages movie subtitles for question generation. For answering these questions, we employ techniques KG and LLM to process questions and retrieve corresponding answers. The proposed method achieves good performance in Deep Video Understanding (DVU) task of NIST TRECVID, demonstrating its effectiveness. © 2023 IEEE.

关键词： Motion pictures

来源：评论

学校读者我要写书评

暂无评论

KAnoCLIP: Zero-Shot Anomaly Detection through Knowledge-Driven Prompt Learning and Enhanced Cross-Modal Integration

KAnoCLIP: Zero-Shot Anomaly Detection through Knowledge-Driv...

引用

2025 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2025

作者： Li, Chengyuan Zhou, Suyang Kong, Jieping Qi, Lei Xue, Hui College of Software Engineering Southeast University Nanjing China School of Computer Science and Engineering Southeast University Nanjing China

ISBN: (纸本)9798350368741

Zero-shot anomaly detection (ZSAD) identifies anomalies without needing training samples from the target dataset, essential for scenarios with privacy concerns or limited data. Vision-language models like CLIP show potential in ZSAD but have limitations: relying on manually crafted fixed textual descriptions or anomaly prompts is time-consuming and prone to semantic ambiguity, and CLIP struggles with pixel-level anomaly segmentation, focusing more on global semantics than local details. To address these limitations, We introduce KAnoCLIP, a novel ZSAD framework that leverages vision-language models. KAnoCLIP combines general knowledge from a Large Language Model (GPT-3.5) and fine-grained, image-specific knowledge from a Visual Question Answering system (Llama3) via Knowledge-Driven Prompt Learning (KnPL). KnPL uses a knowledge-driven (KD) loss function to create learnable anomaly prompts, removing the need for fixed text prompts and enhancing generalization. KAnoCLIP includes the CLIP visual encoder with V-V attention (CLIP-VV), Bi-Directional Cross-Attention for Multi-Level Cross-Modal Interaction (Bi-CMCI), and Conv-Adapter. These components preserve local visual semantics, improve local cross-modal fusion, and align global visual features with textual information, enhancing pixel-level anomaly detection. KAnoCLIP achieves state-of-the-art performance in ZSAD across 12 industrial and medical datasets, demonstrating superior generalization compared to existing methods. © 2025 IEEE.

关键词： Prompt Learning Vision-Language Models Zero-shot Anomaly Detection

来源：评论

学校读者我要写书评

暂无评论

LMR-CBT: learning modality-fused representations with CB-Transformer for multimodal emotion recognition from unaligned multimodal sequences

引用

Frontiers of computer science 2024年第4期18卷 39-47页

作者： Ziwang FU Feng LIU Qing XU Xiangling FU Jiayin QI School of Computer Science(National Pilot Software Engineering School) Beijing University of Posts and TelecommunicationsBeijing 100876China Key Laboratory of Trustworthy Distributed Computing and Service(BUPT) Beijing 100876China Shanghai International School of Chief Technology Officer East China Normal UniversityShanghai 200062China School of Computer Science and Technology East China Normal UniversityShanghai 200062China School of Cyberspace Security Guangzhou UniversityGuangdong 510006China

Learning modality-fused representations and processing unaligned multimodal sequences are meaningful and challenging in multimodal emotion *** approaches use directional pairwise attention or a message hub to fuse language,visual,and audio ***,these fusion methods are often quadratic in complexity with respect to the modal sequence length,bring redundant information and are not *** this paper,we propose an efficient neural network to learn modality-fused representations with CB-Transformer(LMR-CBT)for multimodal emotion recognition from unaligned multi-modal ***,we first perform feature extraction for the three modalities respectively to obtain the local structure of the ***,we design an innovative asymmetric transformer with cross-modal blocks(CB-Transformer)that enables complementary learning of different modalities,mainly divided into local temporal learning,cross-modal feature fusion and global self-attention *** addition,we splice the fused features with the original features to classify the emotions of the ***,we conduct word-aligned and unaligned experiments on three challenging datasets,IEMOCAP,CMU-MOSI,and *** experimental results show the superiority and efficiency of our proposed method in both *** with the mainstream methods,our approach reaches the state-of-the-art with a minimum number of parameters.

关键词： modality-fused representations cross-model blocks multimodal emotion recognition unaligned multimodal sequences computational affection

来源：评论

学校读者我要写书评

暂无评论

Exploration and Practice of Deep Learning Classroom Teaching of 'computer Network' Course Based on the Integration of Specialization and Innovation 5

Exploration and Practice of Deep Learning Classroom Teaching...

引用

5th Asia Conference on Information engineering, ACIE 2025

作者： Huang, Min Peng, ShaoWu Yuan, Hua School of Software Engineering South China University of Technology Guangzhou China School of Computer Science South China University of Technology Guangzhou China

ISBN: (纸本)9798331528409

In the era of rapid development of artificial intelligence technologies, traditional teaching models are unable to meet the employment needs of enterprises, and talent cultivation in universities faces more challenges, so teachers should consider how to better cultivate students' innovative abilities in classroom teaching and guiding them in practical applications. The paper focuses on teaching reform and practice of 'computer Networks' course by deep learning classroom, which mainly includes three aspects: 1) Integrating ideological and political education and innovative education into the process of professional education of in the curriculum, by which to enhance students' professional innovation awareness;2) Inviting experts from enterprises to participate in classroom teaching and practical guidance to help students learn about the ability requirements and related positions in the IT enterprises for talents in the era of AI;3) Designing practice projects and guiding students to implement them by 'design thinking' methods, and also adopting an evaluation method that can reflect the innovative ability of students. Works of the paper can better reflect the support and deepening the teaching philosophy of studentcentered, and provide an attempt to improve teaching effectiveness and talent cultivation quality. © 2025 IEEE.

关键词： Cultivation

来源：评论

学校读者我要写书评

暂无评论

An Approach to Optimize Symbolic Execution in Ethereum Smart Contracts 24

An Approach to Optimize Symbolic Execution in Ethereum Smart...

引用

24th IEEE International Conference on software Quality, Reliability and Security Companion, QRS-C 2024

作者： Liu, Yi Cai, Lizhi Chai, Chenchen Ma, Yuan School of Information Science and Engineering East China University of Science and Technology Shanghai China Shanghai Development Center of Computer Software Technology Shanghai Key Laboratory of Computer Software Testing & Evaluating Shanghai China

ISBN: (纸本)9798350365658

As the application of smart contracts in blockchain technology becomes increasingly widespread, their security issues have emerged as a focal point of both research and practice. Although symbolic execution technology has been extensively applied for the detection of security vulnerabilities in smart contracts, its limitations in terms of execution efficiency and code coverage remain significant. In response to these challenges, this study introduces an innovative method named HoneyToken-DetectorRev, which optimizes traditional symbolic execution through program slicing technology, aiming to enhance the efficiency and effectiveness of smart contract security detection. By selecting ten smart contracts from the Ethereum platform, which vary in business logic and complexity, we conducted an empirical analysis of HoneyToken-DetectorRev. The experimental results demonstrate that this method significantly improves the efficiency of smart contract detection without sacrificing precision. This research not only provides an effective technical approach for the security detection of smart contracts but also offers theoretical and practical guidance for further enhancing the security of blockchain applications. © 2024 IEEE.

关键词： Smart contract

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：