检索结果-内蒙古大学图书馆

11th International Conference on Learning Representations, ICLR 2023

作者： Liu, Zhenghao Xiong, Chenyan Lv, Yuanhuiyi Liu, Zhiyuan Yu, Ge Department of Computer Science and Technology Northeastern University Shenyang China Microsoft Research Redmond United States Department of Computer Science and Technology Tsinghua University Beijing China Institute for Artificial Intelligence Tsinghua University Beijing China State Key Lab on Intelligent Technology and Systems Tsinghua University Beijing China

This paper presents Universal Vision-Language Dense Retrieval (UniVL-DR), which builds a unified model for multi-modal retrieval. UniVL-DR encodes queries and multi-modality resources in an embedding space for searching candidates from different modalities. To learn a unified embedding space for multi-modal retrieval, UniVL-DR proposes two techniques: 1) Universal embedding optimization strategy, which contrastively optimizes the embedding space using the modality-balanced hard negatives;2) Image verbalization method, which bridges the modality gap between images and texts in the raw data space. UniVL-DR achieves the state-of-the-art on the multi-modal open-domain question answering benchmark, WebQA, and outperforms all retrieval models on the two subtasks, text-text retrieval and text-image retrieval. It demonstrates that universal multi-modal search is feasible to replace the divide-and-conquer pipeline with a united model and also benefits single/cross modality tasks. All source codes of this work are available at https://***/OpenMatch/UniVL-DR. © 2023 11th International Conference on Learning Representations, ICLR 2023. All rights reserved.

关键词： Embeddings

来源：评论

学校读者我要写书评

暂无评论

Vision-language navigation with energy-based policy 24

Vision-language navigation with energy-based policy

引用

Proceedings of the 38th International Conference on Neural Information Processing systems

作者： Rui Liu Wenguan Wang Yi Yang The State Key Lab of Brain-Machine Intelligence and College of Computer Science and Technology Zhejiang University Hangzhou China College of Computer Science and Technology Zhejiang University Hangzhou China

ISBN: (纸本)9798331314385

Vision-language navigation (VLN) requires an agent to execute actions following human instructions. Existing VLN models are optimized through expert demonstrations by supervised behavioural cloning or incorporating manual reward engineering. While straightforward, these efforts overlook the accumulation of errors in the Markov decision process, and struggle to match the distribution of the expert policy. Going beyond this, we propose an Energy-based Navigation Policy (ENP) to model the joint state-action distribution using an energy-based model. At each step, low energy values correspond to the state-action pairs that the expert is most likely to perform, and vice versa. Theoretically, the optimization objective is equivalent to minimizing the forward divergence between the occupancy measure of the expert and ours. Consequently, ENP learns to globally align with the expert policy by maximizing the likelihood of the actions and modeling the dynamics of the navigation states in a collaborative manner. With a variety of VLN architectures, ENP achieves promising performances on R2R, REVERIE, RxR, and R2R-CE, unleashing the power of existing VLN models.

关键词：

来源：评论

学校读者我要写书评

暂无评论

TCP:Textual-based Class-aware Prompt tuning for Visual-Language Model

arXiv

引用

arXiv 2023年

作者： Yao, Hantao Zhang, Rui Xu, Changsheng State Key Laboratory of Multimodal Artificial Intelligence Systems Institute of Automation CAS State Key Lab of Processors Institute of Computing Technology CAS China

Prompt tuning represents a valuable technique for adapting pre-trained visual-language models (VLM) to various downstream tasks. Recent advancements in CoOp-based methods propose a set of learnable domain-shared or image-conditional textual tokens to facilitate the generation of task-specific textual classifiers. However, those textual tokens have a limited generalization ability regarding unseen domains, as they cannot dynamically adjust to the distribution of testing classes. To tackle this issue, we present a novel Textual-based Class-aware Prompt tuning(TCP) that explicitly incorporates prior knowledge about classes to enhance their discriminability. The critical concept of TCP involves leveraging Textual Knowledge Embedding (TKE) to map the high generalizability of class-level textual knowledge into class-aware textual tokens. By seamlessly integrating these class-aware prompts into the Text Encoder, a dynamic class-aware classifier is generated to enhance discriminability for unseen domains. During inference, TKE dynamically generates class-aware prompts related to the unseen classes. Comprehensive evaluations demonstrate that TKE serves as a plug-and-play module effortlessly combinable with existing methods. Furthermore, TCP consistently achieves superior performance while demanding less training ***:https://***/htyao89/Textual-based_Class-aware_ prompt_tuning/ Copyright © 2023, The Authors. All rights reserved.

关键词： Visual languages

来源：评论

学校读者我要写书评

暂无评论

Prototype Guided Personalized Federated Intrusion Detection system 10

Prototype Guided Personalized Federated Intrusion Detection ...

引用

10th IEEE Smart World Congress, SWC 2024

作者： Cheng, Long Yan, Huiru Zhou, Hanlin Wang, Ying Tang, Haichuan Fang, Fang North China Electric Power University State Key Laboratory of Alternate Electrical Power System with Renewable Energy Sources School of Control and Computer Engineering Beijing China Institute of Computing Technology Chinese Academy of Sciences Beijing China Ai Lab of Crrc Academy Beijing China

ISBN: (纸本)9798331520861

Preventing network attacks and protecting user privacy are consistently hot research topics in the Internet of Things (IoT) and edge computing fields. Recent advancements in Federated Learning (FL) have shown promise in addressing these challenges. FL allows various clients to collaboratively build an intrusion detection system (IDS) without sharing their private data. However, most existing methods train a single intrusion detection model for all clients and assume that the training and test data distribution are identical, leading to unsatisfactory detection accuracy and generalization abilities in practice. To overcome these challenges, this paper introduces a prototype-guided personalized FL approach named PG-FedIDS. We propose two novel mechanisms within this method. Firstly, we utilize class prototypes as auxiliary information carriers to generate personalized models for each client, rather than generating a single global model as in previous works. Secondly, we propose a prototype-guided ensemble learning strategy, which can leverage the global knowledge in prototypes to enhance detection accuracy and generalization abilities for each client. We conduct extensive experiments on two benchmark datasets with different evaluation test settings. The results demonstrate that our PG-FedIDS achieves promising detection accuracy and consistently outperforms other FL baselines. © 2024 IEEE.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Study on the Relationship Between Space Charge and Corona Initiation Characteristics Around DC Wire 11th

Study on the Relationship Between Space Charge and Corona In...

引用

11th Frontier Academic Forum of Electrical Engineering, FAFEE 2024

作者： Wu, Haitao Wang, Penghao Zhao, Tian Wang, Yuxiang He, Gaohui Liu, Hongbo State Grid Chongqing Electric Power Company Electric Power Science Research Institute Chongqing404100 China China Electric Power Planning & Engineering Institute Xicheng District Beijing100120 China State Key Lab of Reliability and Intelligence of Electrical Equipment School of Electrical Engineering Hebei University of Technology Tianjin300401 China Chongqing400715 China

ISBN: (纸本)9789819788279

In the case of the continuous development of UHV engineering, the field intensity and spatial charge in gas are important parameters to reflect the electromagnetic environment of transmission lines. The study of the relevance between field intensity and spatial charge is not only helpful to better understand the mechanism of corona discharge, but also can provide support for the development and construction of UHV engineering. In this paper, A device that incorporates the ultrasonic pulse method is constructed to measure the distribution of spatial charge in corona cage. At the same time, the corona formation characteristics and the distribution law of electric field intensity are obtained through theoretical calculation and experimental measurement, and then the correlation mechanism between charge and initial electrical field intensity is established. The results show that when corona discharge occurs, the charge density in the cage has positive correlation with the initial voltage and negatively correlated with the initial electric field. The above research results can more accurately reveal the correlation mechanism between the spatial charge and the field intensity, which has crucial role for the stable operation of UHV engineering. © Beijing Paike Culture Commu. Co., Ltd. 2025.

关键词： Electric corona

来源：评论

学校读者我要写书评

暂无评论

Simulation Study on the Morphology of Streamer Propagation in Insulating Oil

Simulation Study on the Morphology of Streamer Propagation i...

引用

International Conference on High Voltage Engineering and Application, ICHVE

作者： Haochen Zuo Zhengyong Huang Jian Li Kai Zhang Chaofan Wang Jawad Ahmad State Key Lab of Power Transmission Equipment & System Security and New Technology School of Electrical Engineering Chongqing University Chongqing China

ISBN: (数字)9798350374988

ISBN: (纸本)9798350374995

Athermal electric model for the initiation and transmission of streamer leading to electrical breakdown has been proposed. We have constructed a migration rate jump mechanism for positive ions, negative ions, and electrons, expanding the macroscopic size of the streamer in insulating oil. The analysis using COMSOL Multiphysics shows that compared to the original model, the shape of the streamer has been greatly expanded, which can better reflect the actual shape of the streamer. In addition, the dynamic development process of the streamer indicates that the initial shape of the streamer does not exhibit branching structures, and only when the electric field threshold is reached, the streamer will exhibit more obvious branching structures. We confirmed the effectiveness of the model in predicting the streamer morphology under different voltages by simulating the streamer morphology under different voltages.

关键词： Shape Oils Simulation Morphology High-voltage techniques Predictive models Ions

来源：评论

学校读者我要写书评

暂无评论

Decoding Respiratory Signals from LFP in Olfactory Bulb based on ConvLSTM Framework 7

Decoding Respiratory Signals from LFP in Olfactory Bulb base...

引用

7th International Conference on Biological Information and Biomedical Engineering, BIBE 2024

作者： Xu, Haoze Yu, Bowen Yu, Chaonan Xu, Kedi Qiushi Academy for Advanced Studies Zhejiang Provincial Key Laboratory of Cardio-Cerebral Vascular Detection Technology and Medicinal Effectiveness Appraisal Key Laboratory for Biomedical Engineering of Ministry of Education School of Biomedical Engineering and Instrument Science Zhejiang University China Nanhu Brain-computer interface institute Hangzhou311100 China State Key Lab of Brain-Machine Intelligence Zhejiang University Hangzhou China

ISBN: (纸本)9783800762910

Respiratory signals are crucial for decoding olfactory neural circuits and understanding respiration-entrained local field potential (LFP) rhythms. Acquiring high-precision respiratory signals often requires additional equipment, increasing the experimental burden, particularly in free-moving paradigms. However, within the current neuroscience paradigm, we have not yet found a non-invasive method for obtaining the respiratory signals of laboratory animals. To minimize surgical trauma to experimental animals, this paper introduces a ConvLSTM-based decoding method, which uses the LFP from the olfactory bulb (OB) region of animals to accurately decode respiratory signals. After validating the effectiveness of the ConvLSTM method through comparative experiments, we further investigated the factors influencing the model's decoding performance. The Gamma frequency band exhibited the best decoding results, with a mean absolute error (MAE) of 0.16±0.09, which may be related to the causal direction from respiratory signals to Gamma rhythms. Additionally, the ConvLSTM model demonstrated strong trend prediction capabilities in multi-sequence point predictions and forward predictions, despite a slight decline in single-point prediction accuracy. Extensive validation on publicly available datasets indicates that the ConvLSTM-based respiratory decoding approach is highly applicable, offering a non-invasive alternative for respiratory signal acquisition in animal studies. © VDE VERLAG GMBH.

关键词： Noninvasive medical procedures

来源：评论

学校读者我要写书评

暂无评论

AWNet: Negative Obstacle Semantic Segmentation Network Based on Adaptive Weight Loss

AWNet: Negative Obstacle Semantic Segmentation Network Based...

引用

IEEE International Conference on Industrial Informatics (INDIN)

作者： Jiarui Zhao Guizhen Yu Jie Wang Zhangyu Wang Junjie Zhang Yunsong Feng State Key Lab of Intelligent Transportation System School of Transportation Science and Engineering The Key Laboratory of Autonomous Transportation Technology for Special Vehicles Ministry of Industry and Information Technology Beihang University Beijing China State Key Lab of Intelligent Transportation System Research Institute for Frontier Science Hefei innovation research institute of Beihang university Beihang University Hefei China State Key Laboratory of Pulsed Power Laser Technology National University of Defense Technology Heufei China

ISBN: (数字)9798331527471

ISBN: (纸本)9798331527488

On roads, both vehicles and pedestrians will leave traces after passing by. Slight traces will not cause any trouble to the passing vehicles and pedestrians subsequently. However, there are certain potholes in the road that have been formed by repeated crushing of vehicles. It may cause mechanical damage to following vehicles and lead to accidents as vehicles get stuck in the potholes. Thus, the segmentation of negative obstacles is crucial for the safe driving of self-driving vehicles. This paper proposed an AWNet neural network to this end, and quantitative analyses were conducted based on the publicly available dataset DRNO. Specifically, a random cut-and-joint module (RCJ) was proposed to augment the dataset. The crossentropy loss function was modified by adding an adaptive weight module, i.e., the adaptive weight cross-entropy loss function. It allows the model to focus on negative obstacles during implementation. Furthermore, this paper proposed a post-processing module for adaptive viewpoint conversion based on depth maps, which is capable of fusing depth maps and RGB images to convert the front view into a BEV viewpoint. The mAcc for Negative obstacle semantic segmentation can reach $78.39\%$ , the mIoU can reach $73.21{{\% }}$ , the mF 1 can reach $79.23{{\% }}$ . The experimental results reveal that the AWNet is able to state-of-the-art performance compared with other networks.

关键词： Adaptation models Pedestrians Statistical analysis Roads Semantic segmentation Semantics Neural networks Data augmentation Data models Informatics

来源：评论

学校读者我要写书评

暂无评论

SpeCache: Speculative key-Value Caching for Efficient Generation of LLMs

arXiv

引用

arXiv 2025年

作者： Jie, Shibo Tang, Yehui Han, Kai Deng, Zhi-Hong Han, Jing State Key Laboratory of General Artificial Intelligence School of Intelligence Science and Technology Peking University China Huawei Noah’s Ark Lab Hong Kong School of Physics Peking University China

Transformer-based large language models (LLMs) have already achieved remarkable results on long-text tasks, but the limited GPU memory (VRAM) resources struggle to accommodate the linearly growing demand for key-value (KV) cache as the sequence length increases, which has become a bottleneck for the application of LLMs on long sequences. Existing KV cache compression methods include eviction, merging, or quantization of the KV cache to reduce its size. However, compression results in irreversible information forgetting, potentially affecting the accuracy of subsequent decoding. In this paper, we propose SPECACHE, which takes full advantage of the large and easily expandable CPU memory to offload the complete KV cache, and dynamically fetches KV pairs back in each decoding step based on their importance measured by low-bit KV cache copy in VRAM. To avoid inference latency caused by CPU-GPU communication, SPECACHE speculatively predicts the KV pairs that the next token might attend to, allowing us to prefetch them before the next decoding step which enables parallelization of prefetching and computation. Experiments on LongBench and Needle-in-a-Haystack benchmarks verify that SPECACHE effectively reduces VRAM usage while avoiding information forgetting for long sequences without re-training, even with a 10× high KV cache compression ratio. © 2025, CC0.

关键词： Decoding

来源：评论

学校读者我要写书评

暂无评论

Progress in quantitative research on the relationship between atmospheric oxidation and air quality

引用

Journal of Environmental Sciences 2023年第1期123卷 350-366页

作者： Yuesi Wang Xin Jin Zirui Liu Gehui Wang Guiqian Tang Keding Lu Bo Hu Shanshan Wang Guohui Li Xinqin An Chao Wang Qihou Hu Lingyan He Fenfen Zhang Yuanhang Zhang State Key Laboratory of Atmospheric Boundary Layer Physics and Atmospheric Chemistry(LAPC) Institute of Atmospheric PhysicsChinese Academy of SciencesBeijing 100029China College of Earth and Planetary Sciences University of Chinese Academy of SciencesBeijing 100049China 3 State Key Joint Laboratory or Environmental Simul ation and Pollution Control College of Environmental Sciences and EngineeringPeking UniversityBeijing 100871China State Key Joint Laboratory of Environmental Simulation and Pollution Control School of EnvironmentTsinghua UniversityBeijing 100084China Key Lab of Geophysical Information System of the Ministry of Education School of Geographic SciencesEast China Normal UniversityShanghai 200241China Shanghai Key Laboratory of Atmospheric Particle Pollution and Prevention(LAP3) Department of Environmental Science and EngineeringFudan UniversityShanghai 201203China Key Lab of Aerosol Chemistry and Physics State Key Laboratory of Loess and Quaternary GeologyInstitute of Earth EnvironmentChinese Academy of SciencesXi’an 710061China Institute of Atmospheric Composition and Environmental Meteorology Chinese Academy of Meteorological SciencesBeijing 100081China Key Lab of Environmental Optics and Technology Anhui Institute of Optics and Fine MechanicsHefei Institutes of Physical ScienceChinese Academy of SciencesHefei 230031China

Atmospheric oxidizing capacity(AOC)is an essential driving force of troposphere chemistry and self-cleaning,but the definition of AOC and its quantitative representation remain *** by national demand for air pollution control in recent years,Chinese scholars have carried out studies on theories of atmospheric chemistry and have made considerable progress in AOC *** paper will give a brief review of these ***,AOC indexes were established that represent apparent atmospheric oxidizing ability(AOIe)and potential atmospheric oxidizing ability(AOIp)based on aspects of macrothermodynamics and microdynamics,respectively.A closed study refined the quantitative contributions of heterogeneous chemistry to AOC in Beijing,and these AOC methods were further applied in Beijing-Tianjin-Hebei and key areas across the *** addition,the detection of ground or vertical profiles for atmospheric OH·,HO_(2)·,NO_(3)·radicals and reservoir molecules can now be obtained with domestic instruments in diverse ***,laboratory smoke chamber simulations revealed heterogeneous processes involving reactions of O_(3)and NO_(2),which are typical oxidants in the surface/interface atmosphere,and the evolutionary and budgetary implications of atmospheric oxidants reacting under multispecies,multiphase and multi-interface conditions were ***,based on the GRAPES-CUACE adjoint model improved by Chinese scholars,simulations of key substances affecting atmospheric oxidation and secondary organic and inorganic aerosol formation have been *** numerical simulations of AOIe and AOIp were performed,and regional coordination of AOC was *** optimized plan for controlling O_(3)and PM2.5was analyzed by scenario simulation.

关键词： Atmospheric oxidizing capacity(AOC) Quantitative index Free radicals Observation and modeling Air quality

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：