检索结果-内蒙古大学图书馆

arXiv 2024年

作者： He, Haorui Song, Yuchen Wang, Yuancheng Li, Haoyang Zhang, Xueyao Wang, Li Huang, Gongping Chng, Eng Siong Wu, Zhizheng School of Data Science Chinese University of H ong Kong Shenzhen518172 China School of Computer Science and Engineering Nanyang Technological University Singapore639798 Singapore School of Electronic Information Wuhan University Wuhan430072 China

One-shot voice conversion (VC) aims to alter the timbre of speech from a source speaker to match that of a target speaker using just a single reference speech from the target, while preserving the semantic content of the original source speech. Despite advancements in one-shot VC, its effectiveness decreases in real-world scenarios where reference speeches, often sourced from the internet, contain various disturbances like background noise. To address this issue, we introduce Noro, a Noise Robust One-shot VC system. Noro features innovative components tailored for VC using noisy reference speeches, including a dual-branch reference encoding module and a noise-agnostic contrastive speaker loss. Experimental results demonstrate that Noro outperforms our baseline system in both clean and noisy scenarios, highlighting its efficacy for real-world applications. Additionally, we investigate the hidden speaker representation capabilities of our baseline system by repurposing its reference encoder as a speaker encoder. The results shows that it is competitive with several advanced self-supervised learning models for speaker representation under the SUPERB settings, highlighting the potential for advancing speaker representation learning through one-shot VC task. © 2024, CC BY.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

Channel-wise attention based binarization for CNN Architectures

Channel-wise attention based binarization for CNN Architectu...

引用

Emerging Systems and Intelligent Computing (ESIC), International Conference on

作者： Ipsita Paul Mainak Bandyopadhyay Satya Ranjan Dash Tanjim Taharat Aurpa School of Computer Engineering KIIT D.U. Bhubaneswar India School of Computer Applications KIIT D.U. Bhubaneswar India Department of Data Science Engineering BSMR Digital University Bangladesh

ISBN: (数字)9798331522100

ISBN: (纸本)9798331522117

Attention mechanisms are known to effectively increase the performance of a model by focusing on specific sections of the input while executing a task. Squeeze and Excitation module (SE-Net) and Efficient channel attention module (ECA-Net) are well-known Channel Attention networks that only involve a few parameters to achieve a clear performance gain. In order to further improve the model’s accuracy, Binarized Neural Networks (BNNs) can be introduced in the same. BNNs are deep learning models that use binarized values for activation functions and training weights, instead of fully precised values that leads to much faster computation, occupying less memory and power. Motivated by this, a method of binarization of the convolutional network is proposed in this paper that can enhance the overall model accuracy. In our work, we focused on evaluating the performance of our binarized attention module over the pre-existing ones. Thus, the objective of this work is to design, implement, and evaluate efficient neural network models that balance accuracy and computational efficiency for image classification tasks on datasets with differing complexity levels, such as CIFAR-10 and CIFAR-100. By exploring various backbone architectures (e.g., MobileNetV2 and ResNet50), attention mechanisms (e.g., ECANet and SENet), and binarization techniques, this study aims to understand the impact of these components on model performance and resource requirements. Ultimately, the goal is to develop optimized models that achieve high accuracy with minimal computational overhead, making them suitable for deployment in resource-constrained environments.

关键词： Deep learning Training Attention mechanisms Accuracy Computational modeling Neural networks Transfer learning Convolutional neural networks Image classification Residual neural networks

来源：评论

学校读者我要写书评

暂无评论

Automated Sentiment Analysis for Web-Based Stock and Cryptocurrency News Summarization with Transformer-Based Models

Automated Sentiment Analysis for Web-Based Stock and Cryptoc...

引用

2023 IEEE Asia-Pacific Conference on computer science and data Engineering, CSDE 2023

作者： Hasan, Mehedi Rahman, Md. Tahmid Alavee, Kazi Ahnaf Zillanee, Abu Hasnayen Uddin, Jia Alam, Md. Golam Rabiul Brac University Department of Computer Science and Engineering Dhaka1212 Bangladesh Brac Business School Brac University 66 Mohakhali Dhaka1212 Bangladesh Endicott College Woosong University Ai and Big Data Department Daejeon Korea Republic of

ISBN: (纸本)9798350341072

In the fast-paced realm of global financial markets, characterized by rapid trading of both stocks and cryptocurren-cies, it has become essential to grasp the influence of sentiment on market dynamics. With more than 630,000 publicly traded companies worldwide and major stock exchanges like the NYSE handling a substantial portion of global equity transactions, the inherent volatility of the stock market is well-established. Over the past decade, various factors have contributed to the consistent fluctuations in stock prices. One key factor is the influence of investor reviews sourced from diverse news outlets and social media platforms such as Twitter. Understanding how these reviews can be collected and effectively summarized is crucial. This paper centers on the intricate field of market sentiment analysis and its profound impact on user sentiment, subsequently affecting price fluctuations in both stocks and cryptocurrencies. In this study, we present a comprehensive exploration of the development and evaluation of an automated sentiment analysis system tailored for summarizing web-based news related to stocks and *** have implemented BERT (Bidirectional Encoder Representations from Transformers) in combination with NLTK for text summarization, a highly accurate model with a performance level of 95.84%, as part of our proposed approach. © 2023 IEEE.

关键词： Sentiment analysis

来源：评论

学校读者我要写书评

暂无评论

Leveraging Attention to Achieve Generalization for Image Forgery Detection

Leveraging Attention to Achieve Generalization for Image For...

引用

Intelligent Methods, Systems, and Applications (IMSA)

作者： Mohamed Atta School of Information Technology and Computer Science Artificial Intelligence Program Nile University Giza Egypt

Recently, image forgery has become an alarming trend with the growth of available easy-to-use editing and generation tools. Modern DeepFake methods have achieved extraordinary progress in realistic face manipulation, thus raising concerns among the public about the misuse of such technologies. Unfortunately, with the obnoxiously wide range of possible manipulation and artifact-covering methods, most existing state-of-the-art detection methods lack the generalization capability to handle the output variations. To address this issue, a noticeable shift towards using attention mechanisms has emerged using balanced portions of the latest challenging datasets to detect intra-and inter-spatial relations. Our paper provides a comprehensive analysis of modern deep learning-based methods, showing the benefits of the shift. In addition, we make propositions for future research directions and dataset-building methodology.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Joint Deployment of UAV and Edge Server In Edge Computing

Joint Deployment of UAV and Edge Server In Edge Computing

引用

2025 IEEE Wireless Communications and Networking Conference, WCNC 2025

作者： Huang, Wei Zhang, Deyu Luo, Wei Tang, Yin School of Computer Science and Engineering Central South University Changsha410083 China Big Data Institute Central South University Changsha410083 China

ISBN: (纸本)9798350368369

Edge server placement is a hot issue in mobile edge computing. It is a key prerequisite for deploying edge servers that can meet computing needs and improve resource utilization. This paper studies the joint location deployment problem of edge server and UAV, and proposes a dynamic placement solution for heterogeneous servers based on dynamically changing network environments and computing resource requirements, with the goal of minimizing migration costs. Extensive comparative experiments based on real datasets demonstrate that our algorithm outperforms other algorithms in minimizing costs. © 2025 IEEE.

关键词： Edge computing

来源：评论

学校读者我要写书评

暂无评论

Ur-Sam: Enhancing the Reliability of Segment Anything Model For Auto-Prompting Medical Image Segmentation with Uncertainty Rectification

SSRN

引用

SSRN 2024年

作者： Zhang, Yichi Hu, Shiyao Ren, Sijie Pan, Tan Jiang, Chen Cheng, Yuan Qi, Yuan Artificial Intelligence Innovation and Incubation Institute Fudan University Shanghai China Shanghai Academy of Artificial Intelligence for Science Shanghai China School of Data Science Fudan University Shanghai China School of Computer Science and Technology Xi’an Jiaotong University Xi’an China

Recent advancements in prompt-driven image segmentation exemplified by the Segment Anything Model (SAM) have shown remarkable potential for universal medical image segmentation. However, their reliance on manual prompting each target structure from testing images poses a significant challenge due to the need for domain-specific knowledge from medical experts and directly increases the burden for applications. Auto-prompting methods have been introduced to enable automatic segmentation but often lack reliability, while low-quality prompts induced by noisy annotation can significantly compromise the accuracy of segmentation. In this paper, we propose UR-SAM, an uncertainty rectified SAM framework to enhance the reliability of auto-prompting medical image segmentation. Building upon a localization framework for automatic prompt generation, we incorporates a prompt augmentation module to obtain a series of input prompts for uncertainty estimation. Subsequently, we adapt an uncertainty-based rectification strategy to leverage the distribution of estimated uncertainty and further improve the segmentation performance without the need of supplementary training or fine-tuning. We conduct extensive experiments on two representative medical datasets covering the segmentation of 22 head-and-neck organs and 13 abdominal organs. Experimental results demonstrate significant improvements in dice similarity coefficient with up to 10.7 % and 13.8 %, demonstrating efficiency and broad capabilities for medical image segmentation without manual prompting. © 2024, The Authors. All rights reserved.

关键词： Reliability

来源：评论

学校读者我要写书评

暂无评论

HORIZON-FREE REGRET FOR LINEAR MARKOV DECISION PROCESSES

arXiv

引用

arXiv 2024年

作者： Zhang, Zihan Lee, Jason D. Chen, Yuxin Du, Simon S. Department of Electrical and Computer Engineering Princeton University United States Department of Statistics and Data Science University of Pennsylvania United States Paul G. Allen School of Computer Science and Engineering University of Washington United States

A recent line of works showed regret bounds in reinforcement learning (RL) can be (nearly) independent of planning horizon, a.k.a. the horizon-free bounds. However, these regret bounds only apply to settings where a polynomial dependency on the size of transition model is allowed, such as tabular Markov Decision Process (MDP) and linear *** give the first horizon-free bound for the popular linear MDP setting where the size of the transition model can be exponentially large or even uncountable. In contrast to prior works which explicitly estimate the transition model and compute the inhomogeneous value functions at different time steps, we directly estimate the value functions and confidence sets. We obtain the horizon-free bound by: (1) maintaining multiple weighted least square estimators for the value functions;and (2) a structural lemma which shows the maximal total variation of the inhomogeneous value functions is bounded by a polynomial factor of the feature dimension. Copyright © 2024, The Authors. All rights reserved.

关键词： Markov processes

来源：评论

学校读者我要写书评

暂无评论

Self-supervised Learning of Orc-Bert Augmentor for Recognizing Few-Shot Oracle Characters 15th

Self-supervised Learning of Orc-Bert Augmentor for Recognizi...

引用

15th Asian Conference on computer Vision, ACCV 2020

作者： Han, Wenhui Ren, Xinlin Lin, Hangyu Fu, Yanwei Xue, Xiangyang School of Data Science Computer Science and MOE Frontiers Center for Brain Science Shanghai Key Lab of Intelligent Information Processing Fudan University Shanghai China

ISBN: (纸本)9783030695439

This paper studies the recognition of oracle character, the earliest known hieroglyphs in China. Essentially, oracle character recognition suffers from the problem of data limitation and imbalance. Recognizing the oracle characters of extremely limited samples, naturally, should be taken as the few-shot learning task. Different from the standard few-shot learning setting, our model has only access to large-scale unlabeled source Chinese characters and few labeled oracle characters. In such a setting, meta-based or metric-based few-shot methods are failed to be efficiently trained on source unlabeled data;and thus the only possible methodologies are self-supervised learning and data augmentation. Unfortunately, the conventional geometric augmentation always performs the same global transformations to all samples in pixel format, without considering the diversity of each part within a sample. Moreover, to the best of our knowledge, there is no effective self-supervised learning method for few-shot learning. To this end, this paper integrates the idea of self-supervised learning in data augmentation. And we propose a novel data augmentation approach, named Orc-Bert Augmentor pre-trained by self-supervised learning, for few-shot oracle character recognition. Specifically, Orc-Bert Augmentor leverages a self-supervised BERT model pre-trained on large unlabeled Chinese characters datasets to generate sample-wise augmented samples. Given a masked input in vector format, Orc-Bert Augmentor can recover it and then output a pixel format image as augmented data. Different mask proportion brings diverse reconstructed output. Concatenated with Gaussian noise, the model further performs point-wise displacement to improve diversity. Experimentally, we collect two large-scale datasets of oracle characters and other Chinese ancient characters for few-shot oracle character recognition and Orc-Bert Augmentor pre-training. Extensive experiments on few-shot learning demonstrate the effe

关键词： Character recognition

来源：评论

学校读者我要写书评

暂无评论

A Blackbox Fuzzing Based on Automated State Machine Extraction

A Blackbox Fuzzing Based on Automated State Machine Extracti...

引用

IEEE International Conference on data science in Cyberspace (DSC)

作者： Zhaowei Zhang Wenjing Yu Zibin Wang Youlin Xiang National Key Laboratory of Science and Technology on Information System Security Beijing China Shanghai Key Laboratory of Data Science School of Computer Science Fudan University Shanghai China

Currently, protocol fuzzing techniques mainly employ two approaches: greybox fuzzing based on mutation and blackbox fuzzing based on generation. Greybox fuzzing techniques use message exchanges between the protocol server and actual clients as seeds, and generate test cases through mutation. Although this approach can provide coverage information of the SUT's code and state space through instrumentation and feedback, its drawback lies in the relatively random mutation strategy, which makes it challenging to validate the SUT's message verification process. This paper addresses this limitation by utilizing artificial intelligence techniques to extract protocol state machines. It aims to overcome the reliance on manual work in blackbox fuzzing based on generation and leverage its advantages to generate more effective fuzzing test cases. The study utilizes Prompt-Learning technology to analyze the semantic information in protocol RFC documents, obtain corresponding intermediate representations, and extract protocol state machines from these representations. Taking the BGP protocol as an experimental subject, the experimental results demonstrate a certain level of accuracy in the obtained protocol state machines and the ability to generate test cases from these state machines, thereby enhancing the automation level of protocol fuzzing.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Two-Level Graph Representation Learning with Community-as-a-Node Graphs

Two-Level Graph Representation Learning with Community-as-a-...

引用

IEEE International Conference on data Mining (ICDM)

作者： Jeong-Ha Park Kisung Lee Hyuk-Yoon Kwon Graduate School of Data Science Seoul National University of Science and Technology Seoul South Korea Division of Computer Science and Engineering Louisiana State University Baton Rouge USA

In this paper, we propose a novel graph representation learning (GRL) model that aims to improve both representation accuracy and learning efficiency. We design a Two-Level GRL architecture based on the graph partitioning: 1) local GRL on nodes within each partitioned subgraph and 2) global GRL on subgraphs. By partitioning the graph through community detection, we enable elaborate node learning in the same community. Based on Two-Level GRL, we introduce an abstracted graph, Community-as-a-Node Graph(CaaN), to effectively maintain the high-level structure with a significantly reduced graph. By applying the CaaN graph to local and global GRL, we propose Two-Level GRL with Community-as-a-Node (CaaN 2L) that effectively maintains the global structure of the entire graph while accurately representing the nodes in each community. A salient point of the proposed model is that it can be applied to any existing GRL model by adopting it as the base model for local and global GRL. Through extensive experiments employing seven popular GRL models, we show that our model outperforms them in both accuracy and efficiency.

关键词：

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：