检索结果-内蒙古大学图书馆

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Thrampoulidis, Christos Department of Electrical and Computer Engineering University of British Columbia Vancouver Canada

We initiate an investigation into the optimization properties of next-token prediction (NTP), the dominant training paradigm for modern language models. Specifically, we study the structural properties of the solutions selected by gradient-based optimizers among the many possible minimizers of the NTP objective. By framing NTP as cross-entropy minimization across distinct contexts, each tied with a sparse conditional probability distribution across a finite vocabulary of tokens, we introduce "NTP-separability conditions" that enable reaching the data-entropy lower bound. With this setup, and focusing on linear models with fixed context embeddings, we characterize the optimization bias of gradient descent (GD): Within the data subspace defined by the sparsity patterns of distinct contexts, GD selects parameters that equate the logits' differences of in-support tokens to their log-odds. In the orthogonal subspace, the GD parameters diverge in norm and select the direction that maximizes a margin specific to NTP. These findings extend previous research on implicit bias in one-hot classification to the NTP setting, highlighting key differences and prompting further research into the optimization and generalization properties of NTP, irrespective of the specific architecture used to generate the context embeddings. © 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

A Review of Modern Computing Paradigms: Cloud, Fog, and Edge

A Review of Modern Computing Paradigms: Cloud, Fog, and Edge

引用

2025 International Conference on Multi-Agent Systems for Collaborative Intelligence, ICMSCI 2025

作者： Sharma, Neha Choudhary, Monika Srivastava, Prakash Prabha, Chander Sharma, Himanshu Cgc Department of Computer Science and Engineering Punjab Landran India Department of Computer Science & Engineering Dehradun India Chitkara University Institute of Engineering and Technology Chitkara University Punjab India

ISBN: (纸本)9798331509828

In the current competitive environment, businesses are looking at how to lower their IT expenses while simultaneously improving their productivity and flexibility to respond quickly to changes in their operational procedures. One of the most widely accepted information technology-based services is cloud computing, which has garnered interest from both the academic and business communities as well as the general public-characteristics such as pay-per-use and subscription options ease of use, elasticity, and low entry costs. Low entry cost in fog, cloud, and edge computing is achieved by utilizing pay-as-you-go pricing models, leveraging open-source technologies, and optimizing resource allocation. By emphasizing how this technology differs from the rest of the current technology and concentrating on its numerous benefits and areas of application-which provide proof of its growing popularity-this paper presents the development of modern computer paradigms and associated areas of research interest. A discussion of each paradigm is done by clearly defining its main ideas and how they relate to the others. The paper presents a comparative assessment of Cloud, Edge, and Fog computing indicating their diverse areas of applicability. © 2025 IEEE.

关键词： Fog computing

来源：评论

学校读者我要写书评

暂无评论

Single-mode waveguides in barium titanate-on-insulator platform fabricated by wet etching

Single-mode waveguides in barium titanate-on-insulator platf...

引用

CLEO: science and Innovations in CLEO 2024, CLEO: S and I 2024 - Part of Conference on Lasers and Electro-Optics

作者： Lin, Hong-Lin Aashna, Pragati Cao, Yu Danner, Aaron Department of Electrical and Computer Engineering National University of Singapore 119077 Singapore

We present, for the first time, single-mode single-crystal barium titanate waveguides on insulator, achieved through a wet-etching method. This etching approach opens new possibilities for the development of barium-ti... 详细信息

关键词： Barium titanate

来源：评论

学校读者我要写书评

暂无评论

Enhancing Image Quality of Aging Smartphones Using Multi-scale Selective Kernel Feature Fusion Network 3rd

Enhancing Image Quality of Aging Smartphones Using Multi-sca...

引用

3rd International Conference on Intelligent Systems and Pattern Recognition, ISPR 2023

作者： Hossain, Md. Yearat Rakib, Md. Mahbub Hasan Nijhum, Ifran Rahman Rahman, Tanzilur Department of Electrical and Computer Engineering North South University Dhaka Bangladesh

ISBN: (纸本)9783031463341

Smartphones are now the most popular medium for photography. Smartphones can capture better images than the hardware they use thanks to Machine Learning-based computational photography. Even though modern phones have sophisticated image enhancement applications, older and aged phones lag behind for a variety of reasons. Lens quality degradation, which results in washed out, soft-looking images, is one of the most common issues seen in older devices. We propose a method for improving such images by using an attention-based multi-scale residual neural network trained on a synthetic dataset. We chose two smartphones: an old device that captures degraded images and a modern flagship that provides reference enhanced images. Then we used the bloom filter, contrast, and highlight adjustment to make the reference images appear degraded. Later, we trained the model using synthetically degraded images and tested it on a variety of older devices. We achieved a maximum Peak Signal to Noise Ratio (PSNR) score of 74 throughout our experiments. To evaluate the model's images, we used Blind Image Quality Assessment (BIQA) methods such as HyperIQA. Aside from correcting the lens issue, the model achieves comparatively better sharpness, contrast, and color processing. Our proposed method generalizes well and achieves up to 11.45% improvement on novel devices by utilizing a very limited amount of data. © 2024, The Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Smartphones

来源：评论

学校读者我要写书评

暂无评论

Sentiment Analysis of Twitter Tweets for COVID- 19 Pfizer Vaccines 7

Sentiment Analysis of Twitter Tweets for COVID- 19 Pfizer Va...

引用

7th International Conference on Intelligent Computing and Control Systems, ICICCS 2023

作者： Shah, Milind Vasant, Avani Gandhi, Kinjal Panesar, Sonia Department of Computer Science & Engineering Gujarat Vadodara India Baroda University School of Computer Science Engineering & Technology Department of Computer Science & Engineering Gujarat Vadodara India

ISBN: (纸本)9798350397253

Coronavirus pandemic, also referred to as COVID-19, broke out in 2019 and caused extensive illnesses. Many experts in the medical field are of the opinion that the epidemic first appeared in Wuhan, China, before it spread to other areas of the world. Washing one's hands, covering one's face, and socially isolating oneself from others are only some of the preventative methods that decision-makers have advocated. Early in 2021, a number of nations, notably the United States, started giving their citizens the COVID-19 vaccination. Since the World Health organization proclaimed that COVID-19 was causing a pandemic, there has been a concerted effort on a worldwide scale to create vaccinations against the virus. Because people often use social media, and Twitter in particular, to share their knowledge and attitudes on global topics such as COVID-19-vaccination, Twitter is an excellent tool for researching public perspectives. This research study performs a sentiment analysis of Pfizer vaccines on the Twitter platform. The dataset contains recent tweets about Pfizer and BioNTech vaccines. In the context of this research, we have investigated people's sentiments regarding vaccinations manufactured by Pfizer. Additionally, the dataset required to build this machine -learning model may be downloaded from Kaggle. The tweets acquired from Twitter have been examined using machine learning algorithms to determine the various sentiments that are included within the tweets. The results show 40.9 % of Tweets with positive segments, while 48.9 % with negative and 10.2% seemed neutral. In terms of model performance evaluation, this study has used different ML classification algorithms to see which classifier gives the best model accuracy. © 2023 IEEE.

关键词： COVID-19

来源：评论

学校读者我要写书评

暂无评论

EEG Signal Denoising Using Beta-Variational Autoencoder 46

EEG Signal Denoising Using Beta-Variational Autoencoder

引用

46th Annual International Conference of the IEEE engineering in Medicine and Biology Society, EMBC 2024

作者： Mahaseni, Behzad Khan, Naimul Mefraz Toronto Metropolitan University Department of Electrical and Computer Engineering Toronto Canada

ISBN: (纸本)9798350371499

Electroencephalography (EEG) signals are a valuable source of information for investigating brain activity and different types of brain-related disease diagnoses. However, EEG signals are often contaminated by various kinds of noises/artifacts. Several methods have been proposed for EEG reconstruction/denoising to facilitate signal analysis, but such algorithms often fail when the EEG contains extreme artifacts. This paper presents a novel method for reconstructing EEG signals using a variant of the variational autoencoder (VAE) called beta-VAE. Through extensive evaluation of our model on the DEAP dataset, we show that the β-VAE architecture learns a compressed representation of the EEG signal in an unsupervised manner, and the reconstructed signal contains less artifact. We compare our proposed method with different baselines and state-of-the-art techniques for EEG signal denoising, demonstrating significantly reduced reconstruction error under artificially induced noise. The results suggest that our approach has great potential for improving the analysis and understanding of EEG signals in clinical and research settings. © 2024 IEEE.

关键词： Electroencephalography

来源：评论

学校读者我要写书评

暂无评论

Fake News Encoder Classifier (FNEC) for Online Published News Related to COVID-19 Vaccines

引用

Intelligent Automation & Soft Computing 2023年第7期37卷 73-90页

作者： Asma Qaiser Saman Hina Abdul Karim Kazi Saad Ahmed Raheela Asif Department of Computer Science and Information Technology NED University of Engineering and TechnologyKarachi75270Pakistan Department of Computer Science Iqra UniversityKarachi76400Pakistan Department of Software Engineering NED University of Engineering and TechnologyKarachi75270Pakistan

In the past few years,social media and online news platforms have played an essential role in distributing news content *** of the authenticity of news has become a major *** the COVID-19 outbreak,misinformation and fake news were major sources of confusion and insecurity among the general *** the first quarter of the year 2020,around 800 people died due to fake news relevant to *** major goal of this research was to discover the best learning model for achieving high accuracy and performance.A novel case study of the Fake News Classification using ELECTRA model,which achieved 85.11%accuracy score,is thus reported in this *** addition to that,a new novel dataset called COVAX-Reality containing COVID-19 vaccine-related news has been *** the COVAX-Reality dataset,the performance of FNEC is compared to several traditional learning models i.e.,Support Vector Machine(SVM),Naive Bayes(NB),Passive Aggressive Classifier(PAC),Long Short-Term Memory(LSTM),Bi-directional LSTM(Bi-LSTM)and Bi-directional Encoder Representations from Transformers(BERT).For the evaluation of FNEC,standard metrics(Precision,Recall,Accuracy,and F1-Score)were utilized.

关键词： Deep learning fake news detection machine learning transformer model classification

来源：评论

学校读者我要写书评

暂无评论

Resource Allocation in Cooperative NOMA: Opportunities and Challenges 3

Resource Allocation in Cooperative NOMA: Opportunities and C...

引用

3rd International Conference on Emerging Trends in electrical, Control, and Telecommunication engineering, ETECTE 2024

作者： Mushtaq, Uzma Baig, Sobia Department of Electrical and Computer Engineering COMSATS University Islamabad Lahore Pakistan

ISBN: (纸本)9798331504861

The radio access technology keeps users connected to communication networks anywhere and anytime in Fifth Generation and beyond wireless communication systems (b5G). Future radio access technologies incorporate major wireless communication system parameters, like huge data rates, massive connectivity, high energy efficiency, very low latency, etc. Hence, there is a need to bring revolutionary changes in the present radio access technologies (RATs). The employment of 5G and beyond technologies aim to provide better throughput, lower latency, excellent user connectivity and quality of service, and reliability. Non-orthogonal multiple access (NOMA) techniques are employed to allow users to share the same time-frequency blocks. NOMA can support huge data rates, massive connectivity and promises user fairness. The user capacity of NOMA can be enhanced by integrating it with the cooperative relay networks, that provide enhanced spectral efficiency, energy efficiency and user connectivity. This article presents the detailed review of resource allocation techniques in Cooperative NOMA (CNOMA) with discussion about the challenges in implementing CNOMA in multi-user communications system. ©2024 IEEE.

关键词： Cooperative communication

来源：评论

学校读者我要写书评

暂无评论

An Integrated Representation & Compression Scheme Based on Convolutional Autoencoders with 4D-DCT Perceptual Encoding for High Dynamic Range Light Fields 8

An Integrated Representation & Compression Scheme Based on C...

引用

8th IEEE International Conference on Information and Communication Technology, CICT 2024

作者： Khaidem, Sally Sharma, Mansi Indian Institute of Technology Madras Department of Electrical Engineering Chennai India Thapar Institute of Engineering and Technology Department of Computer Science and Engineering Patiala India

ISBN: (纸本)9798331531539

The emerging and existing light field displays are highly capable of realistic presentation of 3D scenes on auto-stereoscopic glasses-free platforms. However, the large size of light field data presents a significant drawback for 3D displays and streaming purposes. When a light field has a high dynamic range (HDR), its size increases drastically. This paper proposes a novel compression algorithm specifically designed for high dynamic range light fields that achieves perceptually lossless compression. The algorithm leverages inter-view and intra-view correlations within the HDR light field by treating it as a four-dimensional volume. This compression technique is based on the 4D-DCT Uniform Colour Space (4DDCT-UCS) algorithm. Further encoding of 4DDCT-UCS processed images using High-Efficiency Video Coding (HEVC) eliminates intra-frame, inter-frame, and intrinsic redundancies in HDR light field data. Our comparison with state-of-the-art coders like JPEG-XL and existing HDR video coding algorithms demonstrates the superior compression performance of the proposed scheme for real-world light fields. © 2024 IEEE.

关键词： Encoding (symbols)

来源：评论

学校读者我要写书评

暂无评论

Learning Enabled Control for Optimal EPO Dosage in Virtual CKD Patients: Case of Bleeding and Missing Dosage 46

Learning Enabled Control for Optimal EPO Dosage in Virtual C...

引用

46th Annual International Conference of the IEEE engineering in Medicine and Biology Society, EMBC 2024

作者： Affan, Affan Inanc, Tamer University of Louisville Electrical and Computer Engineering Department KY United States

ISBN: (纸本)9798350371499

Traditional model-based control methods require predictive models to design control policies. These models often suffer limitations on dimensionality, uncertainty, and unmodeled dynamics. This affects the performance of control policy, especially, personalized drug dosing of erythropoietin (EPO) in chronic patients due to the narrow therapeutic window and highly nonlinear time-variant patient characteristics. It is more challenging in the events of bleeding and missing dosage. This paper investigates model-free Reinforcement Learning (RL) and adaptive model predictive control (AMPC) in the events of bleeding and missing dosages for precise EPO dosing in a virtual setting. In this research work, an AMPC policy and Deep Q-Network(DQN)-RL agent is trained for a virtual chronic kidney disease (CKD) patient to find the optimal EPO dosage to maintain the hemoglobin (Hgb) level between 10-12 g/dl. The simulation results show that AMPC and DQN-RL both can provide optimal results while meeting the constraints, however, DQN-RL is more computationally challenging and demands high *** relevance - This research work provides a framework to help clinicians in decision-making for personalized EPO dose guidance using model-free Reinforcement Learning (RL) and adaptive model predictive control (AMPC) in the events of bleeding and missing dosages. © 2024 IEEE.

关键词： Drug dosage

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：