检索结果-内蒙古大学图书馆

arXiv 2025年

作者： Aydemir, Görkay Cai, Xiongyi Xie, Weidi Güney, Fatma Department of Computer Engineering Koç University Turkey KUIS AI Center School of Artificial Intelligence Shanghai Jiao Tong University China

In this paper, we consider the problem of long-term point tracking, which requires consistent identification of points across multiple frames in a video, despite changes in appearance, lighting, perspective, and occlusions. We target online tracking on a frame-by-frame basis, making it suitable for real-world, streaming scenarios. Specifically, we introduce Track-On, a simple transformer-based model designed for online long-term point tracking. Unlike prior methods that depend on full temporal modeling, our model processes video frames causally without access to future frames, leveraging two memory modules —spatial memory and context memory—to capture temporal information and maintain reliable point tracking over long time horizons. At inference time, it employs patch classification and refinement to identify correspondences and track points with high accuracy. Through extensive experiments, we demonstrate that Track-On sets a new state-of-the-art for online models and delivers superior or competitive results compared to offline approaches on seven datasets, including the TAP-Vid benchmark. Our method offers a robust and scalable solution for real-time tracking in diverse applications. © 2025, CC BY.

关键词： Video streaming

来源：评论

学校读者我要写书评

暂无评论

Improving Adversarial Transferability with Neighbourhood Gradient Information

arXiv

引用

arXiv 2024年

作者： Guo, Haijing Wang, Jiafeng Chen, Zhaoyu Jiang, Kaixun Hong, Lingyi Guo, Pinxue Li, Jinglun Zhang, Wenqiang Shanghai Key Lab of Intelligent Information Processing School of Computer Science Fudan University Shanghai200433 China Shanghai Engineering Research Center of AI & Robotics Academy for Engineering & Technology Fudan University Shanghai200433 China Engineering Research Center of Robotics Ministry of Education Academy for Engineering & Technology Fudan University Shanghai200433 China

Deep neural networks (DNNs) are known to be susceptible to adversarial examples, leading to significant performance degradation. In black-box attack scenarios, a considerable attack performance gap between the surrogate model and the target model persists. This work focuses on enhancing the transferability of adversarial examples to narrow this performance gap. We observe that the gradient information around the clean image, i.e. Neighbourhood Gradient Information, can offer high transferability. Leveraging this, we propose the NGI-Attack, which incorporates Example Backtracking and Multiplex Mask strategies, to use this gradient information and enhance transferability fully. Specifically, we first adopt Example Backtracking to accumulate Neighbourhood Gradient Information as the initial momentum term. Multiplex Mask, which forms a multi-way attack strategy, aims to force the network to focus on non-discriminative regions, which can obtain richer gradient information during only a few iterations. Extensive experiments demonstrate that our approach significantly enhances the adversarial transferability. Especially, when attacking numerous defense models, we achieve an average attack success rate of 95.8%. Notably, our method can plugin with any off-the-shelf algorithm to improve their attack performance without additional time cost. Copyright © 2024, The Authors. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Is Your Image a Good Storyteller? 39

Is Your Image a Good Storyteller?

引用

39th Annual AAai Conference on Artificial Intelligence, AAai 2025

作者： Song, Xiujie Pang, Xiaoyi Tang, Haifeng Wu, Mengyue Zhu, Kenny Q. X-LANCE Lab Department of Computer Science and Engineering MoE Key Lab of Artificial Intelligence AI Institute Shanghai Jiao Tong University Shanghai China China Merchants Bank Credit Card Center Shanghai China University of Texas at Arlington ArlingtonTX United States

ISBN: (纸本)157735897X

Quantifying image complexity at the entity level is straightforward, but the assessment of semantic complexity has been largely overlooked. In fact, there are differences in semantic complexity across images. Images with richer semantics can tell vivid and engaging stories and offer a wide range of application scenarios. For example, the Cookie Theft picture is such a kind of image and is widely used to assess human language and cognitive abilities due to its higher semantic complexity. Additionally, semantically rich images can benefit the development of vision models, as images with limited semantics are becoming less challenging for them. However, such images are scarce, highlighting the need for a greater number of them. For instance, there is a need for more images like Cookie Theft to cater to people from different cultural backgrounds and eras. Assessing semantic complexity requires human experts and empirical evidence. Automatic evaluation of how semantically rich an image will be the first step of mining or generating more images with rich semantics, and benefit human cognitive assessment, Artificial Intelligence, and various other applications. In response, we propose the Image Semantic Assessment (ISA) task to address this problem. We introduce the first ISA dataset and a novel method that leverages language to solve this vision problem. Experiments on our dataset demonstrate the effectiveness of our approach. © 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

Overview of Energy Harvesting and Sustainability in Smart Cities

Overview of Energy Harvesting and Sustainability in Smart Ci...

引用

Advancement in Electronics & Communication engineering (AECE), International Conference on

作者： Ramiz Salama Chadi Altrjman Fadi Al-Turjman Satya Prakash Yadav Satvik Vats Department of Computer Engineering AI and Robotics Institute Research Center for AI and IoT. Near East University Near East Boulevard Mersin 10 Turkiye Department of Chemical Engineering Waterloo University ON Canada Artificial Intelligence Engineering Dept AI and Robotics Institute Near East University Nicosia Mersin10 Turkey Department of Computer Science and Engineering G.L. Bajaj Institute of Technology and A1anagement(GLBITA1) Greater Noida India Department of Computer Science and Engineering Graphic Era Hill University Dehradun India

Sustainability is a key factor in the context of smart cities. In smart cities, promoting sustainable energy is one strategy for achieving sustainability and lowering greenhouse gas emissions. Different energy sources are needed to meet the need for affordable and dependable electrical energy, where the environmental impact typically outweighs the cost of generation. In order to increase sustainability in the electrical industry, smart cities are examined in this article. The methodological strategy for this involved conducting a systematic literature review utilizing the PRISMA procedure. 154 journal articles were examined in-depth for this review. Energy efficiency, renewable energy, and energy and urban planning were the themes that the results were organized into. The results of the study showed that: (a) there is a global academic publication landscape for research on smart cities and energy sustainability; (b) publications are unbalanced when comparing the results of studies on the energy use intensity of different continents to those on the energy sustain ability of smart cities; and (c) there is a strong focus on technology-related issues related to energy sustainability, efficiency, and renewables topics in the literature, but much less attention is given to these issues. Urban and energy authorities are informed by the insights produced, and scholars are given ideas for future research directions.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Toward Task Generalization via Memory Augmentation in Meta-Reinforcement Learning

arXiv

引用

arXiv 2025年

作者： Bao, Kaixi Li, Chenhao As, Yarden Krause, Andreas Hutter, Marco Department of Mechanical and Process Engineering ETH Zurich Switzerland ETH AI Center ETH Zurich Switzerland Department of Computer Science ETH Zurich Switzerland

In reinforcement learning (RL), agents often struggle to perform well on tasks that differ from those encountered during training. This limitation presents a challenge to the broader deployment of RL in diverse and dynamic task settings. In this work, we introduce memory augmentation, a memory-based RL approach to improve task generalization. Our approach leverages task-structured augmentations to simulate plausible out-of-distribution scenarios and incorporates memory mechanisms to enable context-aware policy adaptation. Trained on a predefined set of tasks, our policy demonstrates the ability to generalize to unseen tasks through memory augmentation without requiring additional interactions with the environment. Through extensive simulation experiments and real-world hardware evaluations on legged locomotion tasks, we demonstrate that our approach achieves zero-shot generalization to unseen tasks while maintaining robust in-distribution performance and high sample efficiency. Copyright © 2025, The Authors. All rights reserved.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Tensor MIMO Perfect-Equalizer Design Using New Projection-Based Method and Approximation-Error Analysis

引用

IEEE Systems Journal 2024年第1期18卷 668-679页

作者： Chang, Shih-Yu Wu, Hsiao-Chun Kao, Yi-Chih San Jose State University Department of Applied Data Science San JoseCA95192 United States Louisiana State University School of Electrical Engineering and Computer Science Baton RougeLA70803 United States Yuan Ze University Innovation Center for Ai Applications Chungli32003 Taiwan National Yang Ming Chiao Tung University Network and System Division Information Technology Service Center Hsinchu30010 Taiwan

As real environments are quite complex, multiple-input-multiple-output (MIMO) channels should be characterized by the multidimensional-multirelational models. Tensor linear algebra has been emerging as a powerful and promising tool for such MIMO channel modeling and estimation. However, to the best of our knowledge, there hardly exists any algebraic framework of MIMO channel equalization when the channel is modeled as an arbitrary-dimensional transfer tensor (a tensor whose entries are all rational z-functions). In this work, we first propose a new method based on the projection, namely, 'projection-based method,' to determine whether a given transfer tensor is a full-rank tensor or not and carry out the inversion of an invertible transfer tensor. Our idea is employing the projection subspace of the system transfer tensor to simplify the pertinent inversion process. Second, we design the finite-impulse-response (FIR) realization to approximate the inverse of an arbitrary transfer tensor such that all entries of the approximated inverse transfer tensor are finite-degree polynomials of z. Third, we also analyze the computational and memory complexities required by our proposed new projection-based method for checking the full-rank condition of a given transfer tensor and for undertaking the tensor inversion. Our developed new theories and analysis in this work can be applied to determine the FIR model-orders of an approximated inverse MIMO transfer tensor often encountered in signal processing, control, communications, etc. © 2007-2012 IEEE.

关键词： FIR filters

来源：评论

学校读者我要写书评

暂无评论

What Improves the Generalization of Graph Transformers? A Theoretical Dive into the Self-attention and Positional Encoding 41

What Improves the Generalization of Graph Transformers? A Th...

引用

41st International Conference on Machine Learning, ICML 2024

作者： Li, Hongkang Wang, Meng Ma, Tengfei Liu, Sijia Zhang, Zaixi Chen, Pin-Yu Department of Electrical Computer and System Engineering Rensselaer Polytechnic Institute TroyNY United States Department of Biomedical Informatics Stony Brook University Stony BrookNY United States Department of Computer Science and Engineering Michigan State University East LansingMI United States MIT-IBM Watson AI Lab IBM Research MA United States Department of Computer Science and Engineering University of Science and Technology of China Anhui Hefei China IBM Thomas J. Watson Research Center Yorktown HeightsNY United States

Graph Transformers, which incorporate self-attention and positional encoding, have recently emerged as a powerful architecture for various graph learning tasks. Despite their impressive performance, the complex non-convex interactions across layers and the recursive graph structure have made it challenging to establish a theoretical foundation for learning and generalization. This study introduces the first theoretical investigation of a shallow Graph Transformer for semi-supervised node classification, comprising a self-attention layer with relative positional encoding and a two-layer perceptron. Focusing on a graph data model with discriminative nodes that determine node labels and non-discriminative nodes that are class-irrelevant, we characterize the sample complexity required to achieve a desirable generalization error by training with stochastic gradient descent (SGD). This paper provides the quantitative characterization of the sample complexity and number of iterations for convergence dependent on the fraction of discriminative nodes, the dominant patterns, and the initial model errors. Furthermore, we demonstrate that self-attention and positional encoding enhance generalization by making the attention map sparse and promoting the core neighborhood during training, which explains the superior feature representation of Graph Transformers. Our theoretical results are supported by empirical experiments on synthetic and real-world benchmarks. Copyright 2024 by the author(s)

关键词： Stochastic systems

来源：评论

学校读者我要写书评

暂无评论

Code-switched Text Data Augmentation for Chinese-English Mixed Speech Recognition 11

Code-switched Text Data Augmentation for Chinese-English Mix...

引用

11th IEEE Global Conference on Consumer Electronics, GCCE 2022

作者： Lee, Chung-Ting Wang, Teng-Hui Liang, Kai-Wen Le, Phuong Thi Li, Yung-Hui Wang, Jia-Ching National Central University Department of Computer Science and Information Engineering Taiwan AI Research Center Hon Hai Research Institute New Taipei223011 Taiwan

ISBN: (纸本)9781665492324

Code-switching is a common mode of language expression, which means that two or more languages are used interchangeably in a conversation. At present, the development of such code-switching technology in the field of speech recognition research is still limited by insufficient training corpus of text, which affects the system performance. This paper will use a neural network to train a generator to generate code-switching text to expand the corpus to achieve the purpose of improving the mixed recognition rate of Chinese and English. Our method is to use the Chinese and English texts in the SEAME corpus to train the BERT-BiLSTM-CRF model and use the model to know the code-switching position, generating sentences that conform to the characteristics of this corpus. The experimental results show that the method in this paper has better performance than other methods. © 2022 IEEE.

关键词： Speech recognition

来源：评论

学校读者我要写书评

暂无评论

A Novel Attention-based Explainable Deep Learning Framework Towards Medical Image Classification 7

A Novel Attention-based Explainable Deep Learning Framework ...

引用

7th International Symposium on Innovative Approaches in Smart Technologies, ISAS 2023

作者： Muoka, Gladys W. Yi, Ding Ukwuoma, Chiagoziem Chima Martin, Moindi Duke Aydin, Ahmet Arif Al-Antari, Mugahed A. University of Electronic Science and Technology of China School of Information and Software Engineering Chengdu China Chengdu University of Technology Sichuan College of Nuclear Technology and Automation Engineering Chengdu China Unicaf University Lilongwe Department of Computer Science Lilongwe Malawi Inönü University Faculty of Engineering Department of Computer Engineering Turkey Sejong University College of Software & Convergence Technology Daeyang AI Center Department of Artificial Intelligence Seoul Korea Republic of

ISBN: (纸本)9798350383065

Deep learning applications for medical image classification have shown remarkable promise, particularly incorporating attention-based neural networks. This is particularly relevant in medical imaging, where the integration of Artificial Intelligence assists with various imaging tasks, including classification, segmentation, and detection. Deep learning is revolutionizing medical research and playing a significant role in advancing personalized clinical treatment. However, the lack of interpretability in these models presents a significant obstacle to their adoption in clinical practice. Therefore, there is a growing need for a comprehensive understanding of artificial intelligence systems and their internal mechanisms, capabilities, and limitations, which is the focus of the field of explainable ai. This study proposes a novel attention-based explainable deep learning framework for medical image classification tasks, including Covid-19, breast cancer (BreakHis), lung cancer (LC2500), and Retinal optical coherence tomography (OCT). The proposed framework recorded overall accuracies of 98% (Covid-19 Radiography), 95% (BreakHis), 99.8% (LC2500), and 95%(OCT). For visual analysis of the outcomes, we employ and use the LIME, SHAP, and ELI-5 to analyze the achieved results. The study's primary goal is to bridge the gap between the high performance achieved by attention-based models and the necessity for transparency and interpretability in medical image diagnostics. © 2023 IEEE.

关键词： Optical tomography

来源：评论

学校读者我要写书评

暂无评论

Development of a Taiwanese Speech Synthesis System Using Hidden Markov Models and a Robust Tonal Phoneme Corpus

引用

Journal of Information Science and engineering 2024年第2期40卷 283-302页

作者： Sher, Yung-Ji Hsu, Ming-Chun Chiu, Yu-Hsien Chen, Yeou-Jiunn Wu, Chung-Hsien Wu, Jiunn-Liang Department of Special Education Graduate Institute of Rehabilitation Counseling National Taiwan Normal University Taipei106 Taiwan Wistron Neweb Corporation Hsinchu300 Taiwan NSTC AI Biomedical Research Center National Cheng Kung University Tainan701 Taiwan Department of Computer Science and Information Engineering National Cheng Kung University Tainan701 Taiwan Department of Electrical Engineering Southern Taiwan University of Science and Technology Tainan710 Taiwan Department of Otorhinolaryngology National Cheng Kung University Hospital College of Medicine National Cheng Kung University Tainan704 Taiwan

The number of young native speakers of Taiwanese, the variant of Southern Min spoken in Taiwan, has decreased. Technological advancements such as text-to-speech (TTS) systems could help arrest this decline. The aim of this study was to design a robust tonal phoneme corpus and a speech synthesis system for Modern Literal Taiwanese (MLT). MLT subsyllables were analyzed using phonetics and phonology to establish tonal phoneme models. These robust tonal phoneme models and hidden Markov models were used to construct an MLT TTS synthesis system. Algorithm-based training resulted in 869 balanced sentences containing 12,544 syllables, with each sentence containing an average of 14.4 syllables. In total, 218 sentences, which included rare phonemes, were manually drafted to supplement the corpus. The synthesized phonemes were deemed to have high intelligibility and could be included in the developed TTS system. According to the HTK speech recognition tool, the overall phoneme recognition rate was 96.47%. Testers, who were native Taiwanese speakers, assigned the synthesized sentences a mean opinion score of 4, indicating that they sounded natural. This developed system and the results described herein can inspire future developments in speech technology and computational linguistics. © 2024 Institute of Information Science. All rights reserved.

关键词： Hidden Markov models

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：