检索结果-内蒙古大学图书馆

38th Conference on Neural Information processing Systems, NeurIPS 2024

作者： Zuo, Jialong Nie, Ying Zhou, Hanyu Zhang, Huaxin Wang, Haoyu Guo, Tianyu Sang, Nong Gao, Changxin National Key Laboratory of Multispectral Information Intelligent Processing Technology School of Artificial Intelligence and Automation Huazhong University of Science and Technology China Huawei Noah's Ark Lab Canada

Recent researches have proven that pre-training on large-scale person images extracted from internet videos is an effective way in learning better representations for person re-identification. However, these researches are mostly confined to pre-training at the instance-level or single-video tracklet-level. They ignore the identity-invariance in images of the same person across different videos, which is a key focus in person re-identification. To address this issue, we propose a Cross-video Identity-cOrrelating pre-traiNing (CION) framework. Defining a noise concept that comprehensively considers both intra-identity consistency and inter-identity discrimination, CION seeks the identity correlation from cross-video images by modeling it as a progressive multi-level denoising problem. Furthermore, an identity-guided self-distillation loss is proposed to implement better large-scale pre-training by mining the identity-invariance within person images. We conduct extensive experiments to verify the superiority of our CION in terms of efficiency and performance. CION achieves significantly leading performance with even fewer training samples. For example, compared with the previous state-of-the-art [9], CION with the same ResNet50-IBN achieves higher mAP of 93.3% and 74.3% on Market1501 and MSMT17, while only utilizing 8% training samples. Finally, with CION demonstrating superior model-agnostic ability, we contribute a model zoo named ReIDZoo to meet diverse research and application needs in this field. It contains a series of CION pre-trained models with spanning structures and parameters, totaling 32 models with 10 different structures, including GhostNet, ConvNext, RepViT, FastViT and so on. © 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

Hyperbolic Representations for Prompt Learning 30

Hyperbolic Representations for Prompt Learning

引用

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

作者： Chen, Nan Su, Xiangdong Bao, Feilong College of Computer Science Inner Mongolia University Hohhot China National & Local Joint Engineering Research Center of Intelligent Information Processing Technology for Mongolian China Inner Mongolia Key Laboratory of Mongolian Information Processing Technology China

ISBN: (纸本)9782493814104

Continuous prompt tuning has gained significant attention for its ability to train only continuous prompts while freezing the language model. This approach greatly reduces the training time and storage for downstream tasks. In this work, we delve into the hierarchical relationship between the prompts and downstream text inputs. In prompt learning, the prefix prompt acts as a module to guide the downstream language model, establishing a hierarchical relationship between the prefix prompt and subsequent inputs. Furthermore, we explore the benefits of leveraging hyperbolic space for modeling hierarchical structures. We project representations of pre-trained models from Euclidean space into hyperbolic space using the Poincaré disk which effectively captures the hierarchical relationship between the prompt and input text. The experiments on natural language understanding (NLU) tasks illustrate that hyperbolic space can model the hierarchical relationship between prompt and text input. We release our code at https://***/myaxxxxx/Hyperbolic-Prompt-Learning. © 2024 ELRA Language Resource Association: CC BY-NC 4.0.

关键词： Computational linguistics

来源：评论

学校读者我要写书评

暂无评论

Double-Mix-Net: A Multimodal Music Emotion Recognition Network with Multi-layer Feature Mixing 10th

Double-Mix-Net: A Multimodal Music Emotion Recognition Netwo...

引用

10th Conference on Sound and Music technology, CSMT 2023

作者： Li, Peilin Chen, Kairan Wei, Weixin Zhao, Jiahao Li, Wei School of Computer Science and Technology Fudan University Shanghai Shanghai200438 China Shanghai Key Laboratory of Intelligent Information Processing Fudan University Shanghai Shanghai200433 China

ISBN: (纸本)9789819779611

Computational music emotion recognition (MER) is an important task that aims to recognize emotional content in music tracks. Understanding the emotional content of music can help tailor therapeutic interventions to specific emotional needs, potentially improving their effectiveness. Piano music is widely used in music therapy for depression, anxiety, and other mental health conditions. However, comparatively little work has been done on MER in piano music, which is a challenging task due to the unique difficulties in recognizing emotions in single-timbre music without vocals, especially on valence. In this paper, we propose a novel Double-Mix-Net model that utilizes the multi-layer feature mixing and multi-modal feature mixing methods to improve classification performance. Our model uses both the title of the YouTube music video and the music context to predict the emotion of a given music track. Additionally, we make full use of different granularity features through the multi-layer feature mixing method. Our proposed model achieved state-of-the-art accuracy performance of 73.8% on the Emopia dataset and is the first multi-modal network designed for piano emotion recognition. The results of our study highlight the importance of considering the music title and utilizing multi-layer feature mixing methods in piano music emotion recognition. This work helps the emotional understanding of piano music. Our proposed model can be applied in various domains such as music recommendation systems, film scoring, and game design to enhance the emotional impact of music. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Game design

来源：评论

学校读者我要写书评

暂无评论

Inductive Relation Prediction by Disentangled Subgraph Structure

引用

Tsinghua Science and technology 2024年第5期29卷 1566-1579页

作者： Guiduo Duan Rui Guo Wenlong Luo Guangchun Luo Tianxi Huang Laboratory of Intelligent Collaborative Computing University of Electronic Science and Technology of ChinaChengdu 611731Chinaand also with Trusted Cloud Computing and Big Data Key Laboratory of Sichuan ProvinceChengdu 611731China School of Computer Science and Engineering University of Electronic Science and Technology of ChinaChengdu 611731China Hangzhou NetEase Cloud Music Technology Co. Ltd.Hangzhou 310052China Department of Fundamental Courses Chengdu Textile CollegeChengdu 611700China

Currently,most existing inductive relation prediction approaches are based on subgraph structures,with subgraph features extracted using graph neural networks to predict ***,subgraphs may contain disconnected regions,which usually represent different semantic *** not all semantic information about the regions is helpful in relation prediction,we propose a relation prediction model based on a disentangled subgraph structure and implement a feature updating approach based on relevant semantic *** indirectly achieve the disentangled subgraph structure from a semantic perspective,the mapping of entity features into different semantic spaces and the aggregation of related semantics on each semantic space are *** disentangled model can focus on features having higher semantic relevance in the prediction,thus addressing a problem with existing approaches,which ignore the semantic differences in different subgraph ***,using a gated recurrent neural network,this model enhances the features of entities by sorting them by distance and extracting the path information in the ***,it is shown that when there are numerous disconnected regions in the subgraph,our model outperforms existing mainstream models in terms of both Area Under the Curve-Precision-Recall(AUC-PR)and Hits@*** prove that semantic differences in the knowledge graph can be effectively distinguished and verify the effectiveness of this method.

关键词： disentangled subgraph structure knowledge graph completion Gated Recurrent Unit(GRU) feature updating

来源：评论

学校读者我要写书评

暂无评论

A lightweight object detection network based on YOLOv5 for SAR image 2

A lightweight object detection network based on YOLOv5 for S...

引用

2nd IEEE International Conference on Signal, Information and Data processing, ICSIDP 2024

作者： Shen, Aijia Zhao, Liangbo Xu, Fanyun Wang, Guoqing Liu, Wenchao Shen, Zimeng Beijing Institute of Technology National Key Laboratory of Science and Technology on Space-Born Intelligent Information Processing Beijing China Institute of Remote Sensing Satellite China Academy of Space Technology Beijing China

ISBN: (纸本)9798331515669

Remote sensing image target detection is one of the key technologies in the field of intelligent interpretation of remote sensing images, and it has significant application value in various areas, including military defense. When performing remote sensing image object detection on airborne and spaceborne platforms, the vast amount of remote sensing data processing and limited computational resources impose high real-time requirements on the object detection algorithms. This paper designs a lightweight object detection network model named YOLOv5-tiny, based on the existing deep learning network detection model YOLOv5s, and deploys it on the Jetson TX2 development board for training and testing. Experimental results show that the proposed YOLOv5-tiny model, when tested on the SAR-AIRcraft-1.0 with an input image size of 640x640, is 10 times smaller than YOLOv5s;it has a computational cost of 5.2 GFLOPs, which is 1/5 of YOLOv5s, and the processing time for a single image is reduced to half that of YOLOv5s, with only a 0.1% decrease in accuracy. © 2024 IEEE.

关键词： Object tracking

来源：评论

学校读者我要写书评

暂无评论

Semantic Tagging Algorithm for Tibetan La Case Function Words Oriented to Classification of La Case 4

Semantic Tagging Algorithm for Tibetan La Case Function Word...

引用

4th International Symposium on Computer Engineering and intelligent Communications, ISCEIC 2023

作者： Ban, Mabao Duo, Jiecairang Zhang, Rui Yang, Maojia Cai, Rangjia The State Key Laboratory of Tibetan Intelligent Information Processing and Application Xining China Tibetan Information Processing and Machine Translation Key Laboratory of Qinghai Province Xining China Tibetan Information Processing Engineering Technology and Research Center of Qinghai Province Xining China

ISBN: (纸本)9798350306323

Automatic classification of Tibetan La case () is a challenging and significant task in Tibetan natural language processing. Compared with other short text classification tasks, the automatic classification of La case depend more on the semantic information of context and function words of La case. Based on the definition about La case classification task, this paper designs a semantic labeling algorithm for Tibetan La case function words by analyzing their usage and semantic features. When training La case classification model, calling this algorithm can integrate the semantic information about La case function words, thus increasing the diversity of input features and enriching the semantic information of embedded vectors. Experiments on eight commonly used short text classification models show that the classification accuracy (ACC) after fusing the semantic information on La case function words is improved by 2.1 percentage points on average, which verifies the effectiveness and universality of the algorithm. © 2023 IEEE.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

Optical performance monitoring of mode division multiplexing signal based on multi-task convolutional neural network 7

Optical performance monitoring of mode division multiplexing...

引用

7th Optics Young Scientist Summit, OYSS 2024

作者： He, Ting Chang, Huan Gao, Ran Wang, Fei Guo, Dong Beijing Institute of Technology No.5 Zhongguancun South Street Haidian District Beijing100081 China National Key Laboratory of Science and Technology on Space-Born Intelligent Information Processing Beijing China

ISBN: (数字)9781510688377

ISBN: (纸本)9781510688360

Mode division multiplexing (MDM) optical fiber communication technology enhances bandwidth utilization and communication capacity to address the increasing demand for high-capacity data transmission in the era of rapid network traffic growth. For MDM optical fiber communication systems affected by various impairments, optical performance monitoring (OPM) can effectively characterize the impairments in received optical signals, directly reflecting the transmission signal state and channel conditions, and help ensure efficient and stable network operation. In this paper, we propose a multi-task convolutional neural network (MT-CNN) aided OPM method for MDM signals. The proposed MTCNN combines convolutional neural networks and a multi-task learning framework to extract the transmission link information and predict the optical signal-to-noise ratio (OSNR) of few-mode channels synchronously. Simulation results show that the proposed MT-CNN-based OPM scheme can achieve accurate OSNR performance evaluation for MDM fiber transmission signals. © 2025 SPIE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Prototype-Guided Representation Distribution Optimization for Change Detection in High-Resolution Remote Sensing Imagery 2

Prototype-Guided Representation Distribution Optimization fo...

引用

2nd IEEE International Conference on Signal, Information and Data processing, ICSIDP 2024

作者： Wang, Guoqing Chen, He Wei, Xin Shi, Junqing Chen, Liang Liu, Wenchao Beijing Institute of Technology National Key Laboratory of Science and Technology on Space-Born Intelligent Information Processing Beijing China China Academy of Space Technology Institute of Remote Sensing Satellite Beijing China

ISBN: (纸本)9798331515669

Change detection is the process of automatically extracting change areas from bi-temporal remote sensing images. The key of deep learning-based change detection methods is to extract features that represent changes/non-changes. Most of the existing methods only enhance the representation ability of the changed regions through the model architecture design, but lack further constraints in the representation space. This paper introduces a prototype-guided representation distribution optimization method. By comparing pixel features with prototypes, confusing pixels can be determined, and these pixels are adaptively weighted during learning, so that similar representations are closer and different representations are farther apart. In addition, the representation distribution is also applied as transferred knowledge to knowledge distillation for lightweight CD model. Experimental results on two public datasets show the effectiveness of the proposed method. © 2024 IEEE.

关键词： Change detection

来源：评论

学校读者我要写书评

暂无评论

Global Information Edge Augmentation Network for Remote Sensing Image Continuous Super-Resolution 2

Global Information Edge Augmentation Network for Remote Sens...

引用

2nd IEEE International Conference on Signal, Information and Data processing, ICSIDP 2024

作者： Gao, Shize Lei, Yong Zhang, Jingya Wang, Guoqing Liu, Wenchao Beijing Institute of Technology National Key Laboratory of Science and Technology on Space-Born Intelligent Information Processing Beijing China China Academy of Space Technology Institute of Remote Sensing Satellite Beijing China

ISBN: (纸本)9798331515669

The continuous-scale super-resolution (SR) method for remote sensing (RS) images has the potential to achieve flexible scale SR with a single network, representing a significant research area in the field of RS. However, the majority of prevailing methods learn features from localization, thereby neglecting the global semantic coherence of RS images. This results in unbalanced SR outcomes. Furthermore, the optimization strategy based on the multilayer perceptron (MLP) utilized in the implicit neural representation (INR) method results in blurred reconstructed image edges. To address these issues, a novel continuous-scale SR method for RS images, a global information edge augmentation network (GIEAN), is proposed. Initially, the global state space model (GSSM) aggregates the global information of the image from diverse perspectives and learns the contextual interactions at disparate locations of the image globally. Subsequently, the dual edge enhancement module (DEEM) learns the image body and edges independently through the main branch and edge branch, respectively, with the objective of enhancing the edge component of the SR results. Extensive experimental evaluation on the UCMerced dataset demonstrates the superiority of GIEAN over existing continuous-scale SR techniques. © 2024 IEEE.

关键词： Image enhancement

来源：评论

学校读者我要写书评

暂无评论

Event-Triggered Cooperative Tracking Control of Multiagent Systems with a Dynamic Leader via Approximate Dynamic Programming

IEEE Transactions on Artificial Intelligence

引用

IEEE Transactions on Artificial Intelligence 2024年第6期5卷 2752-2765页

作者： Fu, Hao He, Haodong Chen, Yang Wuhan University of Science and Technology Hubei Province Key Laboratory of Intelligent Information Processing and Real-time Industrial System The School of Computer Science and Technology Wuhan430081 China Wuhan University of Science and Technology Eng. Res. Center for Metallurgical Automation and Measurement Technology of Ministry of Education Wuhan430081 China

This article investigates the optimal cooperative tracking control problem for unknown nonlinear multiagent systems with a real dynamic leader and uncertainties. Its main difficulty lies in eliminating effect of the unmatched uncertainty and noncooperation in the event-triggered manner. To overcome this difficulty, a distributed robust event-triggered control algorithm is developed with the aid of a disturbance observer and robust approximate dynamic programming (ADP). Specifically, this problem is converted into a cooperative multiplayer game for the distributed auxiliary systems employing observer-based feedforward compensation, which contributes to reducing conservatism;via an adaptive event-triggered mechanism and a new weight tuning law, a distributed robust ADP approach is proposed with hope to obviate judging the existence of the saddle point beforehand. Moreover, the local neighbor tracking error and the weight estimation error are demonstrated to be uniformly ultimately bounded by means of Lyapunov stability theory. We also prove that Zeno behavior is excluded. A comparative simulation example for multiquadrotor systems is carried out to verify the effectiveness and superiority of our algorithm. © 2020 IEEE.

关键词： Dynamic programming

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：