检索结果-内蒙古大学图书馆

Enhance Training Objectives for Image Captioning with Decomposed Sequence-level Metric

学校读者我要写书评

暂无评论

Enhance Training Objectives for Image Captioning with Decomp...

International Joint Conference on Neural Networks (IJCNN)

作者： Yehuan Wang Jiefeng Long Jian Hu Lin Shang Department of Computer Science and Technology State Key Laboratory for Novel Software Technology Nanjing University Nanjing China School of Mechanical Engineering Nanjing University of Science and Technology Nanjing China

ISBN: (数字)9798350359312

ISBN: (纸本)9798350359329

Image captioning aims to generate fluent and accurate descriptions for images. To evaluate the quality of captions, various metrics have been proposed. However, current metrics only assess captions at sequence-level, which overwhelms the distinctions between each token. Thus, existing objectives tend to treat each token equally, assigning them with identical weights in loss functions. Intuitively, key words in a caption carry the primary information and contribute more than other words to sequence-level metrics. They should be distinguished and weighted more during training. In this work, we propose to explicitly measure each word and guide the model to focus more on key words in captions. Firstly, we devise token-level CIDEr (CIDEr-T) as a new metric to quantify the importance of each word, by decomposing the sequence-level CIDEr into token-level granularity. CIDEr-T maintains consistency with CIDEr and shows the distinctions between tokens. Thus, we engage CIDEr-T scores of each token as their unique weights in the raw loss functions, which can bridge the gap between training and evaluation.

关键词： Measurement Training Accuracy Neural networks

LEARNING THE DYNAMIC CORRELATIONS AND MITIGATING NOISE BY HIERARCHICAL CONVOLUTION FOR LONG-TERM SEQUENCE FORECASTING

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Yu, Zhihao Ma, Liantao Wang, Yasha Zhao, Junfeng School of Computer Science Peking University Beijing China Key Laboratory of High Confidence Software Technologies Ministry of Education Beijing China National Engineering Research Center of Software Engineering Peking University Beijing China

Deep learning algorithms, especially Transformer-based models, have achieved significant performance by capturing long-range dependencies and historical information. However, the power of convolution has not been fully investigated. Moreover, most existing works ignore the dynamic interaction among variables and evolutionary noise in series. Addressing these issues, we propose a Hierarchical Memorizing Network (HMNet). In particular, a hierarchical convolution structure is introduced to extract the information from the series at various scales. Besides, we propose a dynamic variable interaction module to learn the varying correlation and an adaptive denoising module to search and exploit similar patterns to alleviate noises. These modules can cooperate with the hierarchical structure from the perspective of fine to coarse grain. Experiments on five benchmarks demonstrate that HMNet significantly outperforms the state-of-the-art models by 10.6% on MSE and 5.7% on MAE. Our code is released at https://***/yzhHoward/HMNet. Copyright © 2023, The Authors. All rights reserved.

关键词： Convolution

A Knowledge Acquisition Method for Consumer Electronics Based on Triadic Hesitant Fuzzy Concept

学校读者我要写书评

暂无评论

IEEE Transactions on Consumer Electronics 2025年

作者： Xin, Xianwei Zhao, Ye Wang, Chenyang Xu, Jiucheng Taleb, Tarik Henan Normal University School of Computer and Information Engineering Xinxiang453007 China Key Laboratory of Computational Intelligence and Chinese Information Processing Ministry of Education Taiyuan237016 China Computer and Software Engineering School of College of Shenzhen University Guangdong Shenzhen518060 China Key Laboratory of Artificial Intelligence and Personalized Learning in Education of Henan Province Henan Province 453007 China Ruhr University Bochum Faculty of Electrical Engineering and Information Technology Bochum44801 Germany

With the growing intelligence and personalization of consumer electronic products, extracting effective information has become a key challenge in Consumer Electronics (CE). Triadic Concept Analysis (TCA), an important extension of Formal Concept Analysis (FCA), is widely applied but often struggles to handle the hesitant fuzzy characteristics in real-world data. To address this gap, we propose a Triadic Hesitant Fuzzy Concept (THFC) model. First, we introduce the concept of a triadic hesitant fuzzy background, integrating the fuzzy and hesitant nature of data in the CE environment. We propose a knowledge acquisition method based on THFC, which constructs a multi-granularity knowledge space with a dynamic adjustment mechanism. This method captures subtle distinctions between different types of information and their hierarchical structures in the context of CE, offering significant support for knowledge discovery and decision-making in complex environments. Building on this, we design a variable-precision hybrid similarity measurement (VP-HSM) method to extract core, adjunct, and edge concepts, thus providing more precise and structured knowledge for data analysis and decision-making in CE. Experimental results on publicly available datasets show that the proposed method significantly improves classification accuracy and stability, effectively handling hesitant fuzzy data and improving the accuracy of concept extraction. © 1975-2011 IEEE.

关键词： Data accuracy

Leveraging Predicate and Triplet Learning for Scene Graph Generation

学校读者我要写书评

暂无评论

Leveraging Predicate and Triplet Learning for Scene Graph Ge...

Conference on computer Vision and Pattern Recognition (CVPR)

作者： Jiankai Li Yunhong Wang Xiefan Guo Ruijie Yang Weixin Li IRIP Lab School of Computer Science and Engineering Beihang University Beijing China State Key Laboratory of Complex & Critical Software Environment Beihang University Beijing China Shanghai Artificial Intelligence Laboratory Shanghai China

ISBN: (数字)9798350353006

ISBN: (纸本)9798350353013

Scene Graph Generation (SGG) aims to identify entities and predict the relationship triplets in visual scenes. Given the prevalence of large visual variations of subject-object pairs even in the same predicate, it can be quite challenging to model and refine predicate representations directly across such pairs, which is however a common strategy adopted by most existing SGG methods. We observe that visual variations within the identical triplet are relatively small and certain relation cues are shared in the same type of triplet, which can potentially facilitate the relation learning in SGG. Moreover, for the long-tail problem widely studied in SGG task, it is also crucial to deal with the limited types and quantity of triplets in tail predicates. Accordingly, in this paper, we propose a Dual-granularity Relation Modeling (DRM) network to leverage fine-grained triplet cues besides the coarse-grained predicate ones. DRM utilizes contexts and semantics of predicate and triplet with Dual-granularity Constraints, generating compact and balanced representations from two perspectives to facilitate relation recognition. Furthermore, a Dual-granularity Knowledge Transfer (DKT) strategy is introduced to transfer variation from head predicates/triplets to tail ones, aiming to enrich the pattern diversity of tail classes to alleviate the long-tail problem. Extensive experiments demonstrate the effectiveness of our method, which establishes new state-of-the-art performance on Visual Genome, Open Image, and GQA datasets. Our code is available at https://***/jkli1998/DRM

关键词： Visualization Head Triples (Data structure) Semantics Genomics Tail Pattern recognition

Leveraging Predicate and Triplet Learning for Scene Graph Generation

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Li, Jiankai Wang, Yunhong Guo, Xiefan Yang, Ruijie Li, Weixin IRIP Lab School of Computer Science and Engineering Beihang University Beijing China State Key Laboratory of Complex & Critical Software Environment Beihang University Beijing China Shanghai Artificial Intelligence Laboratory Shanghai China

关键词： Semantics

SFedXL: Semi-Synchronous Federated Learning With Cross-Sharpness and Layer-Freezing

学校读者我要写书评

暂无评论

IEEE Internet of Things Journal 2025年第12期12卷 18692-18707页

作者： Zhao, Mingxiong Zhao, Shihao Feng, Chenyuan Yang, Howard H. Niyato, Dusit Quek, Tony Q. S. Yunnan University National Pilot School of Software Kunming650500 China Ministry of Education Engineering Research Center of Integration and Application of Digital Learning Technology Beijing100039 China Ministry of Education Engineering Research Center of Cyberspace Kunming650504 China Yunnan University of Finance and Economics Yunnan Key Laboratory of Service Computing Kunming650221 China EURECOM Sophia Antipolis06410 France Zhejiang University University of Illinois Urbana-Champaign Institute Zhejiang University Haining314400 China Nanyang Technological University School of Computer Science and Engineering 639798 Singapore Singapore University of Technology and Design Information Systems Technology and Design Pillar 487372 Singapore

Federated learning (FL) emerges as a potential solution for enabling multiple terminal devices to collaboratively accomplish computational tasks within an Unmanned Aerial Vehicle (UAV) swarm. However, traditional FL approaches, predicated on synchronous data aggregation, are not feasible for a UAV swarm owing to the inherently variable and dynamic nature of their communication networks compared with terrestrial systems. Furthermore, the data procured by UAVs is often highly heterogeneous, attributable to disparities in deployment environments and device attributes. Considering the distinct flight paths and unique operational conditions encountered by different UAVs, a considerable amount of data remains unlabeled. To tackle the challenges associated with asynchronous operations and the prevalence of unlabeled data, we introduce a novel framework termed Semi-synchronous FL with Cross-Sharpness and Layer-Freezing (SFedXL), tailored for a UAV swarm. In particular, we devise a cross-sharpness model training strategy aimed at optimizing the utilization of both labeled and unlabeled datasets. Additionally, we propose an innovative semi-synchronous model aggregation protocol, complemented by client-specific layer-freezing and client cluster scheduling, designed to expedite the training process. Our simulation results indicate that the proposed algorithm surpasses current FL methods in terms of object recognition accuracy and communication efficiency, albeit with a trade-off of increased local computation latency. © 2014 IEEE.

关键词： Federated learning

SABER: Model-agnostic Backdoor Attack on Chain-of-Thought in Neural Code Generation

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Jin, Naizhu Li, Zhong Guo, Yinggang Su, Chao Zhang, Tian Zeng, Qingkai State Key Laboratory for Novel Software Technology Nanjing University Jiangsu Nanjing210023 China School of Computer Science and Engineering Changshu Institute of Technology Jiangsu Suzhou215506 China

Recent studies have proposed integrating Chain-of-Thought (CoT) reasoning to further enhance the reliability of Code Language Models (CLMs) in generating code, a step-by-step approach that breaks down complex programming tasks into manageable sub-problems. Advances in this area have introduced CoT models, specifically designed to integrate CoT reasoning effectively into language models, achieving notable improvements in code generation. Despite these advancements, the security of CoT models has not been systematically studied. In this study, we aim to fill this gap by investigating the vulnerability of CoT models to backdoor injection in code generation tasks. To address this, we propose a model-agnostic backdoor attack method SABER (Self-Attention-BasEd backdooR) based on the self-attention mechanism. SABER begins by selecting a malicious output as the backdoor using code mutation operations. It then identifies tokens most relevant to poisoned content by analyzing self-attention scores in the CodeBERT model. Finally, it applies semantic-preserving perturbations to generate adaptive and natural triggers. Our experiments on HumanEval-CoT and OpenEval-CoT test sets demonstrate that CoT models are susceptible to backdoor attacks via data poisoning. Taking the OpenEval-CoT dataset as an example, SABER achieves an ASR of 76.19%, representing an improvement of 14.29% over RIPPLe and a substantial 23.08% enhancement compared to BadPre. Further evaluations using ONION for automated detection and human studies reveal that SABER is stealthier and harder to detect, bypassing 77.27% of automated detection, with a human detection rate of just 3.17%. Our findings reveal that backdoors can be injected into CoT models to manipulate downstream code generation tasks. This highlights the urgent need for further research to understand and mitigate the security vulnerabilities in CoT models. We advocate for increased efforts to explore the security threats facing CoT models and to devel

关键词： Semantics

An Improved Pigeon-Inspired Optimization for Multi-focus Noisy Image Fusion

学校读者我要写书评

暂无评论

Journal of Bionic engineering 2021年第6期18卷 1452-1462页

作者： Yingda Lyu Yunqi Zhang Haipeng Chen Public Computer Education and Research Center Jilin UniversityChangchun130012China College of Software Jilin UniversityChangchun 130012China College of Computer Science and Technology Jilin UniversityChangchun 130012China Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education Jilin UniversityChangchun 130012China

Image fusion technology is the basis of computer vision task,but information is easily affected by noise during *** this paper,an Improved Pigeon-Inspired Optimization(IPIO)is proposed,and used for multi-focus noisy image fusion by combining with the boundary handling of the convolutional sparse *** two-scale image decomposition,the input image is decomposed into base layer and detail *** the base layer,IPIO algorithm is used to obtain the optimized weights for fusion,whose value range is gained by fusing the edge ***,the global information entropy is used as the fitness index of the IPIO,which has high efficiency especially for discrete optimization *** the detail layer,the fusion of its coefficients is completed by performing boundary processing when solving the convolution sparse representation in the frequency *** sum of the above base and detail layers is as the final fused *** results show that the proposed algorithm has a better fusion effect compared with the recent algorithms.

关键词： Improved pigeon-inspired optimization Convolutional sparse representation Noisy image fusion Bionic algorithm

Academically intelligent LLMs are not necessarily socially intelligent

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Xu, Ruoxi Lin, Hongyu Han, Xianpei Sun, Le Sun, Yingfei School of Electronic Electrical and Communication Engineering University of Chinese Academy of Sciences Beijing China Chinese Information Processing Laboratory Institute of Software Chinese Academy of Sciences Beijing China State Key Laboratory of Computer Science Institute of Software Chinese Academy of Sciences Beijing China

The academic intelligence of large language models (LLMs) has made remarkable progress in recent times, but their social intelligence performance remains unclear. Inspired by established human social intelligence frameworks, particularly Daniel Goleman’s social intelligence theory, we have developed a standardized social intelligence test based on real-world social scenarios to comprehensively assess the social intelligence of LLMs, termed as the Situational Evaluation of Social Intelligence (SESI). We conducted an extensive evaluation with 13 recent popular and state-of-art LLM agents on SESI. The results indicate the social intelligence of LLMs still has significant room for improvement, with superficially friendliness as a primary reason for errors. Moreover, there exists a relatively low correlation between the social intelligence and academic intelligence exhibited by LLMs, suggesting that social intelligence is distinct from academic intelligence for LLMs. Additionally, while it is observed that LLMs can’t "understand" what social intelligence is, their social intelligence, similar to that of humans, is influenced by social factors. © 2024, CC BY.

关键词： Computational linguistics