检索结果-内蒙古大学图书馆

39th Annual AAAI Conference on Artificial Intelligence, AAAI 2025

作者： Lan, Pengxiang Yang, Enneng Liu, Yuting Guo, Guibing Zhao, Jianzhe Wang, Xingwei Software College Northeastern University China School of Computer Science and Engineering Northeastern University China

ISBN: (纸本)157735897X

Prompt tuning is a promising method to fine-tune a pre-trained language model without retraining its large-scale parameters. Instead, it attaches a soft prompt to the input text, whereby downstream tasks can be well adapted by merely learning the embeddings of prompt tokens. Nevertheless, existing methods still suffer from two challenges: (i) they are hard to balance accuracy and efficiency. A longer (shorter) soft prompt generally leads to a better (worse) accuracy but at the cost of more (less) training time. (ii) The performance may not be consistent when adapting to different downstream tasks. We attribute it to the same embedding space but responsible for different requirements of downstream tasks. To address these issues, we propose an Efficient Prompt Tuning method (EPT) by multi-space projection and prompt fusion. Specifically, it decomposes a given soft prompt into a shorter prompt and two low-rank matrices, significantly reducing the training time. Accuracy is also enhanced by leveraging low-rank matrices and the short prompt as additional knowledge sources to enrich the semantics of the original short prompt. In addition, we project the soft prompt into multiple subspaces to improve the performance consistency, and then adaptively learn the combination weights of different spaces through a gating network. Experiments on 13 natural language processing downstream tasks show that our method significantly and consistently outperforms 11 comparison methods with the relative percentage of improvements up to 12.9%, and training time decreased by 14%. Copyright © 2025, Association for the Advancement of Artificial Intelligence (***). All rights reserved.

关键词： Learning to rank

来源：评论

学校读者我要写书评

暂无评论

Augmenting Sequential Recommendation with Balanced Relevance and Diversity 39

Augmenting Sequential Recommendation with Balanced Relevance...

引用

39th Annual AAAI Conference on Artificial Intelligence, AAAI 2025

作者： Dang, Yizhou Zhang, Jiahui Liu, Yuting Yang, Enneng Liang, Yuliang Guo, Guibing Zhao, Jianzhe Wang, Xingwei Software College Northeastern University China School of Computer Science and Engineering Northeastern University China

ISBN: (纸本)157735897X

By generating new yet effective data, data augmentation has become a promising method to mitigate the data sparsity problem in sequential recommendation. Existing works focus on augmenting the original data but rarely explore the issue of imbalanced relevance and diversity for augmented data, leading to semantic drift problems or limited performance improvements. In this paper, we propose a novel Balanced data Augmentation Plugin for Sequential Recommendation (BASRec) to generate data that balance relevance and diversity. BASRec consists of two modules: Single-sequence Augmentation and Cross-sequence Augmentation. The former leverages the randomness of the heuristic operators to generate diverse sequences for a single user, after which the diverse and the original sequences are fused at the representation level to obtain relevance. Further, we devise a reweighting strategy to enable the model to learn the preferences based on the two properties adaptively. The Cross-sequence Augmentation performs nonlinear mixing between different sequence representations from two directions. It produces virtual sequence representations that are diverse enough but retain the vital semantics of the original sequences. These two modules enhance the model to discover fine-grained preferences knowledge from single-user and cross-user perspectives. Extensive experiments verify the effectiveness of BASRec. The average improvement is up to 72.0% on GRU4Rec, 33.8% on SASRec, and 68.5% on FMLP-Rec. We demonstrate that BASRec generates data with a better balance between relevance and diversity than existing methods. Copyright © 2025, Association for the Advancement of Artificial Intelligence (***). All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Strong stabilized joint probabilistic data association method of autoregulatory gating in vehicle nonlinear multi-target tracking 4

Strong stabilized joint probabilistic data association metho...

引用

4th International Conference on Mechanical engineering, Intelligent Manufacturing, and Automation Technology, MEMAT 2023

作者： Han, Guoxin Liu, Fuyun Deng, Jucai Liu, Huiqi Bai, Weihua College of Mechanical and Electrical Engineering Guilin University of Electronic Technology Guilin China College of Computer Science and Software Zhaoqing University Zhaoqing China Commercial Vehicle Technology Center Liuzhou Dongfeng Automobile Company Ltd. Liuzhou China

ISBN: (纸本)9781510674820

This paper proposed a strong stabilized joint probabilistic data association approach based on numerical stability optimization and autoregulatory gating strategy for multi-target tracking of vehicles with multi-maneuvering states in a nonlinear cluttered environment. The approach utilized an enhanced square-root cubature Kalman filter to achieve the optimization for joint probabilistic data association, which advanced the accuracy and stability of nonlinear state estimation during the multi-target tracking. On this basis, an autoregulation method of association gate based on tracking error sensitivity was proposed to improve the data association gating strategy, which effectively advanced the success rate of target association and tracking accuracy. The virtual simulation traffic scenes of intelligent driving were constructed by PreScan to better reproduce the multi-target tracking problem. The results of the simulations show that the method has clear advantages in terms of target tracking precision, numerical stability and environmental adaptability. © 2024 SPIE.

关键词： Target tracking

来源：评论

学校读者我要写书评

暂无评论

Efficient and Effective Prompt Tuning via Prompt Decomposition and Compressed Outer Product

arXiv

引用

arXiv 2025年

作者： Lan, Pengxiang Xu, Haoyu Yang, Enneng Liang, Yuliang Guo, Guibing Zhao, Jianzhe Wang, Xingwei Software College Northeastern University China School of Computer Science and Engineering Northeastern University China

Prompt tuning (PT) offers a cost-effective alternative to fine-tuning large-scale pre-trained language models (PLMs), requiring only a few parameters in soft prompt tokens added before the input text. However, existing PT approaches face two significant issues: (i) They overlook intrinsic semantic associations between soft prompt tokens, leading to high discreteness and limited interactions, thus reducing the model’s comprehension and effectiveness in complex tasks. (ii) Due to the complexity of downstream tasks, long soft prompt is necessitated to improve performance, but prompt length correlates positively with memory usage and computational costs. Achieving high efficiency and performance remains an ongoing challenge. To address these issues, we propose a novel Low-pAraMeters Prompt Tuning (LAMP) method, which leverages prompt decomposition and compressed outer product. Specifically, the prompt decomposition module employs Truncated SVD to reduce training parameters and significantly lower the dimensionality of the soft prompt parameter space. It then utilizes a compressed outer product module to facilitate multiple interactions among prompt tokens, exploring their intrinsic associations to enhance knowledge representation. Finally, LAMP uses average pooling to reduce memory usage and training/inference time. Extensive experiments across six architectures and eight datasets demonstrate that LAMP outperforms state-of-the-art PT-based and LoRA-based methods in performance and efficiency. © 2025, CC BY.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

FedDGL: Federated Dynamic Graph Learning for Temporal Evolution and Data Heterogeneity 16

FedDGL: Federated Dynamic Graph Learning for Temporal Evolut...

引用

16th Asian Conference on Machine Learning, ACML 2024

作者： Xie, Zaipeng Li, Likun Chen, Xiangbin Yu, Hao Huang, Qian Key Laboratory of Water Big Data Technology of Ministry of Water Resources Hohai University Nanjing China College of Computer Science and Software Engineering Hohai University Nanjing China College of Artificial Intelligence and Automation Hohai University Nanjing China

Federated graph learning enhances federated learning by enabling privacy-preserving collaborative training on distributed graph data. While traditional methods are effective in managing data heterogeneity, they typically assume static graph structures, overlooking the dynamic nature of real-world graphs. Integrating federated graph learning with dynamic graph neural networks addresses this issue but often fails to retain previously acquired knowledge, limiting generalization for both global and personalized models. This paper proposes FedDGL, a novel framework designed to address temporal evolution, and data heterogeneity in federated dynamic graph learning. Unlike conventional approaches, FedDGL captures temporal dynamics through a global knowledge distillation technique and manages client heterogeneity using a global prototype-based regularization method. The framework employs contrastive learning to generate global prototypes, enhancing feature representation while utilizing a prototype similarity-based personalized aggregation strategy for effective adaptation to local and global data distributions. Experiments on multiple benchmark datasets show that FedDGL achieves significant performance improvements over state-of-the-art methods, with up to 9.02% and 8.77% gains in local and global testing, respectively, compared to FedAvg. These results highlight FedDGL’s effectiveness in improving personalized and global model performance in dynamic, heterogeneous federated graph learning scenarios. © 2024 Z. Xie, L. Li, X. Chen, H. Yu & Q. Huang.

关键词： Dynamic Graph Learning Federated Graph Learning Knowledge Distillation Personalized Aggregation Prototype Learning Temporal Evolution

来源：评论

学校读者我要写书评

暂无评论

Prompting DirectSAM for Semantic Contour Extraction in Remote Sensing Images

Prompting DirectSAM for Semantic Contour Extraction in Remot...

引用

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

作者： Miao, Shiyu Chen, Delong Liu, Fan Zhang, Chuanyi Gu, Yanhui Guo, Shengjie Zhou, Jun College of Computer Science and Software Engineering Hohai University Nanjing China Hong Kong University of Science and Technology Hong Kong School of Computer and Electronic Information Nanjing Normal University Nanjing China School of Information and Communication Technology Griffith University QLD Australia

ISBN: (纸本)9798350368741

The Direct Segment Anything Model (DirectSAM) excels in class-agnostic contour extraction. In this paper, we explore its use by applying it to optical remote sensing imagery, where semantic contour extraction-such as identifying buildings, road networks, and coastlines-holds significant practical value. Those applications are currently handled via training specialized small models separately on small datasets in each domain. We introduce a foundation model derived from DirectSAM, termed DirectSAM-RS, which not only inherits the strong segmentation capability acquired from natural images, but also benefits from a large-scale dataset we created for remote sensing semantic contour extraction. This dataset comprises over 34k image-text-contour triplets, making it at least 30 times larger than individual dataset. DirectSAM-RS integrates a prompter module: a text encoder and cross-attention layers attached to the DirectSAM architecture, which allows flexible conditioning on target class labels or referring expressions. We evaluate the DirectSAM-RS in both zero-shot and fine-tuning setting, and demonstrate that it achieves state-of-the-art performance across several downstream benchmarks. © 2025 IEEE.

关键词： Remote sensing Semantic contour extraction Vision-language learning

来源：评论

学校读者我要写书评

暂无评论

Over time analysis of the codon usage of SARS-CoV-2 and its variants

引用

Computational and Structural Biotechnology Journal 2025年 27卷 2034-2050页

作者： Davidson, Alma Parr, Marina Totzeck, Franziska Churkin, Alexander Barash, Danny Frishman, Dmitrij Tuller, Tamir Department of Biomedical Engineering Tel Aviv University Tel Aviv 6997801 Israel Department of Bioinformatics School of Life Sciences Technical University of Munich Freising Germany Department of Software Engineering Sami Shamoon College of Engineering Beer-Sheva84100 Israel Department of Computer Science Ben-Gurion University Beer-Sheva8410501 Israel Sagol School of Neuroscience Tel Aviv University Tel Aviv 6997801 Israel

As viruses evolve over time with their host, they adapt to multiple cellular functions to ensure efficient long-term transmission. Their ability to survive and function efficiently depends on optimizing their genetic code to effectively recruit the host's gene expression machinery, particularly the translation machinery. Codon usage bias (CUB) measures the level of adaptation at the codon level, considering multiple factors such as the host's tRNA *** estimating the adaptation scores of viruses to their host, we can gain insight into the changes that occur on the genomic level throughout their evolution. In our study, we propose tracking the viral evolution of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) using CUB to estimate adaptation. By comparing different strains and analyzing changes over time, we demonstrate an increased adaptation of the Omicron SARS-COV-2 variant. In addition, we observe fluctuations in CUB scores, with periods of decreased and then increased adaptation over time, offering detailed insights at a fine temporal resolution. This analysis further contributes to understanding how different mutations influence viral adaptation, as well as the evolution of other human-infecting viruses. © 2025 The Authors

关键词： Coronavirus

来源：评论

学校读者我要写书评

暂无评论

A Neuro-Fuzzy Approach to Road Traffic Congestion Prediction

引用

computers, Materials & Continua 2022年第10期73卷 295-310页

作者： Mohammed Gollapalli Atta-ur-Rahman Dhiaa Musleh Nehad Ibrahim Muhammad Adnan Khan Sagheer Abbas Ayesha Atta Muhammad Aftab Khan Mehwash Farooqui Tahir Iqbal Mohammed Salih Ahmed Mohammed Imran BAhmed Dakheel Almoqbil Majd Nabeel Abdullah Omer Department of Computer Information System(CIS) College of Computer Science and Information Technology(CCSIT)Imam Abdulrahman Bin Faisal UniversityDammam31441Saudi Arabia Department of Computer Science(CS) College of Computer Science and Information Technology(CCSIT)Imam Abdulrahman Bin Faisal UniversityDammam31441Saudi Arabia Pattern Recognition and Machine Learning Lab Department of Software EngineeringGachon UniversityInchonKorea Department of Computer Science National College of Business Administration and EconomicsLahorePunjab54000Pakistan Department of Computer Science Government College University(GCU)LahorePunjab54000Pakistan Department of Computer Engineering(CE) College of Computer Science and Information Technology(CCSIT)Imam Abdulrahman Bin Faisal UniversityDammam31441Saudi Arabia College of Business Administration Imam Abdulrahman Bin Faisal UniversityDammam31441Saudi Arabia Department of Networks and Communications(NC) College of Computer Science and Information Technology(CCSIT)Imam Abdulrahman Bin Faisal UniversityDammam31441Saudi Arabia

The fast-paced growth of artificial intelligence applications provides unparalleled opportunities to improve the efficiency of various *** as the transportation sector faces many obstacles following the implementation and integration of different vehicular and environmental aspects *** congestion is among the major issues in this regard which demands serious attention due to the rapid growth in the number of vehicles on the *** address this overwhelming problem,in this article,a cloudbased intelligent road traffic congestion prediction model is proposed that is empowered with a hybrid Neuro-Fuzzy *** aim of the study is to reduce the delay in the queues,the vehicles experience at different road junctions across the *** proposed model also intended to help the automated traffic control systems by minimizing the congestion particularly in a smart city environment where observational data is obtained from various implanted Internet of Things(IoT)sensors across the *** due preprocessing over the cloud server,the proposed approach makes use of this data by incorporating the neuro-fuzzy ***,it possesses a high level of accuracy by means of intelligent decision making with minimum error *** results reveal the accuracy of the proposed model as 98.72%during the validation phase in contrast to the highest accuracies achieved by state-of-the-art techniques in the literature such as 90.6%,95.84%,97.56%and 98.03%,*** far as the training phase analysis is concerned,the proposed scheme exhibits 99.214% accuracy. The proposed prediction modelis a potential contribution towards smart cities environment.

关键词： Neuro-fuzzy machine learning congestion prediction AI cloud computing smart cities

来源：评论

学校读者我要写书评

暂无评论

Automatic Eyewitness Identification During Disasters by Forming a Feature-Word Dictionary

引用

computers, Materials & Continua 2022年第9期72卷 4755-4769页

作者： Shahzad Nazir Muhammad Asif Shahbaz Ahmad Hanan Aljuaid Shahbaz Ahmad Yazeed Ghadi Zubair nawaz Department of Computer Science National Textile UniversityFaisalabadPakistan Computer Sciences Department College of Computer and Information SciencesPrincess Nourah bint Abdulrahman University(PNU)P.O.Box 84428Riyadh 11671Saudi Arabia Department of Software Engineering and Computer Science Al Ain UniversityAbu DhabiUnited Arab Emirates Department of Data Science University of the PunjabPakistan

Social media provide digitally interactional technologies to facilitate information sharing and exchanging ***,in case of disasters,a massive corpus is placed on platforms such as *** accounts can benefit humanitarian organizations and agencies,but identifying the eyewitness Tweets related to the disaster from millions of Tweets is *** approaches have been developed to address this kind of *** recent state-of-the-art system was based on a manually created dictionary and this approach was further refined by introducing linguistic ***,these approaches suffer from limitations as they are dataset-dependent and not *** this paper,we proposed a method to identify eyewitnesses from *** experiment,we utilized 13 features discovered by the pioneer of this domain and can classify the tweets to determine the *** each feature,a dictionary of words was created with the Word Dictionary Maker algorithm,which is the crucial contribution of this *** algorithm inputs some terms relevant to a specific feature for its initialization and then creates the words ***,keyword matching for each feature in tweets is *** a feature exists in a tweet,it is termed as 1;otherwise,***,for 13 features,we created a file that reflects features in each *** classify the tweets based on features,Naïve Bayes,Random Forest,and Neural Network were *** approach was implemented on different disasters like earthquakes,floods,hurricanes,and Forest *** results were compared with the state-of-the-art linguistic rule-based system with 0.81 F-measure *** the same time,the proposed approach gained a 0.88 value of *** results were comparable as the proposed approach is not ***,it can be used for the identification of eyewitness accounts.

关键词： Word dictionary social media eyewitness identification disasters

来源：评论

学校读者我要写书评

暂无评论

View Transformation Robustness for Multi-View 3D Object Reconstruction With Reconstruction Error-Guided View Selection

arXiv

引用

arXiv 2024年

作者： Zhang, Qi Luo, Zhouhang Yu, Tao Huang, Hui College of Computer Science and Software Engineering Shenzhen University Shenzhen China Shenzhen China

View transformation robustness (VTR) is critical for deep-learning-based multi-view 3D object reconstruction models, which indicates the methods’ stability under inputs with various view transformations. However, existing research seldom focuses on view transformation robustness in multi-view 3D object reconstruction. One direct way to improve the models’ VTR is to produce data with more view transformations and add them to model training. Recent progress on large vision models, particularly Stable Diffusion models, has provided great potential for generating 3D models or synthesizing novel view images with only a single image input. To fully utilize the power of Stable Diffusion models without causing extra inference computation burdens, we propose to generate novel views with Stable Diffusion models for better view transformation robustness. Instead of synthesizing random views, we propose a reconstruction error-guided view selection method, which considers the reconstruction errors’ spatial distribution of the 3D predictions and chooses the views that could cover the reconstruction errors as much as possible. The methods are trained and tested on sets with large view transformations to validate the 3D reconstruction models’ robustness to view transformations. Extensive experiments demonstrate that the proposed method can outperform state-of-the-art 3D reconstruction methods and other view transformation robustness comparison methods. Codel will be available at: https://***/zqyq/VTR. © 2024, CC BY-NC-ND.

关键词： 3D reconstruction

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：