检索结果-内蒙古大学图书馆

arXiv 2025年

作者： Singh, Siddharth Singhania, Prajwal Ranjan, Aditya Kirchenbauer, John Geiping, Jonas Wen, Yuxin Jain, Neel Hans, Abhimanyu Shu, Manli Tomar, Aditya Goldstein, Tom Bhatele, Abhinav Department of Computer Science University of Maryland College Park United States ELLIS Institute Tübingen Max Planck Institute for Intelligent Systems Tübingen Germany Department of Electrical Engineering and Computer Sciences University of California Berkeley United States

Training and fine-tuning large language models (LLMs) with hundreds of billions to trillions of parameters requires tens of thousands of GPUs, and a highly scalable software stack. In this work, we present a novel four-dimensional hybrid parallel algorithm implemented in a highly scalable, portable, open-source framework called AxoNN. We describe several performance optimizations in AxoNN to improve matrix multiply kernel performance, overlap non-blocking collectives with computation, and performance modeling to choose performance optimal configurations. These have resulted in unprecedented scaling and peak flop/s (bf16) for training of GPT-style transformer models on Perlmutter (620.1 Petaflop/s), Frontier (1.381 Exaflop/s) and Alps (1.423 Exaflop/s). While the abilities of LLMs improve with the number of trainable parameters, so do privacy and copyright risks caused by memorization of training data, which can cause disclosure of sensitive or private information at inference time. We highlight this side effect of scale through experiments that explore "catastrophic memorization," where models are sufficiently large to memorize training data in a single pass, and present an approach to prevent it. As part of this study, we demonstrate fine-tuning of a 405-billion parameter LLM using AxoNN on Frontier. Copyright © 2025, The Authors. All rights reserved.

关键词： Supercomputers

来源：评论

学校读者我要写书评

暂无评论

Microcombs for ultradense optical communications

Microcombs for ultradense optical communications

引用

Signal Processing in Photonic Communications, SPPCom 2021 - Part of OSA Advanced Photonics Congress 2021

作者： Corcoran, Bill Department of Electrical and Computer Systems Engineering Photonic Communications Lab Monash University ClaytonVIC3800 Australia

Optical microcombs provide a compact and energy-efficient method to generate ultrahigh data rate signals from a single light source. We summarize some recent advances in this field, and some future prospectsv. © ...

ISBN: (纸本)9781557528209

关键词：

来源：评论

学校读者我要写书评

暂无评论

Recurrent Optical Spectrum Slicing Receiver for Power Fading Mitigation in Highly Dispersive Links using Programmable Photonics

Recurrent Optical Spectrum Slicing Receiver for Power Fading...

引用

IEEE Photonics Conference (IPC)

作者： K. Sozos F. Da Ros Metodi Yankov Stavros Deligiannidis G. Sarantoglou C. Mesaritakis A. Bogris Department of Informatics and Computer Engineering University of West Attica Athens Greece Department of Electrical and Photonics Engineering Technical University of Denmark Lyngby Denmark Department of Information and Communication Systems Engineering University of the Aegean Karlovassi Greece

We experimentally validate our works in recurrent optical spectrum slicers for dispersion compensation in IM/DD links, exploiting programmable photonics. We equalize 50 km of 64 Gb/s PAM-4 transmission in C-band, with... 详细信息

ISBN: (数字)9798350361957

ISBN: (纸本)9798350361964

关键词： Fading channels C-band Prevention and mitigation Optical computing Optical receivers Optical fiber communication Dispersion Photonics

来源：评论

学校读者我要写书评

暂无评论

The use of combined remote electronic monitoring and blood concentrations of antipsychotics for assessment of therapeutic adherence in patients with schizophrenia: results of a prospective study

引用

BMC Psychiatry 2025年第1期25卷 1-11页

作者： Šilhán, Petr Hýža, Martin Ambroš, Samuel Dostálek, Petr Schwarzová, Jana Skřont, Tomáš Perničková, Denisa Baarová, Pavla Augustynek, Martin Kacířová, Ivana Uřinovská, Romana Department of Psychiatry University Hospital Ostrava 17. listopadu 1790 Ostrava Poruba 708 52 Czech Republic Department of Clinical Neurosciences Faculty of Medicine University of Ostrava Ostrava Czech Republic Department of Cybernetics and Biomedical Engineering Faculty of Electrical Engineering and Computer Science VSB-Technical University Ostrava Ostrava Czech Republic Department of Clinical Pharmacology Institute of Laboratory Medicine University Hospital Ostrava Ostrava Czech Republic Institute of Clinical Pharmacology Faculty of Medicine University of Ostrava Ostrava Czech Republic

Background: Schizophrenia is a serious mental illness, the pharmacological treatment of which comprises primarily the use of antipsychotics. However, non-adherence to their use and its reliable determination present a serious clinical and economic problem. This study aimed to determine therapeutic adherence in outpatients with schizophrenia spectrum disorders by combining short-term electronic monitoring of dispenser opening with the measurement of antipsychotic blood concentrations. Methods: A total of 55 patients underwent a week-long electronic monitoring of dispenser opening and measurement of blood concentrations before and after monitoring. Patients who correctly opened the dispenser at least in 80% of scheduled time points during the weekly interval and, at the same time, did not show a change in blood concentration of the antipsychotic by more than 30% in any direction, were considered adherent. Results: 69.1% of the patients met the adherence criteria, which was less than that determined by the Drug Attitude Inventory (DAI-10), the Visual Analogue Scale (VAS), and the Clinician Rating Scale (CRS). 7.3% of the patients took less than 80% of the prescribed doses and a change in blood concentrations of the antipsychotic by more than 30% was detected in 25.4% of the patients. In 70.9% of patients, the detected concentrations were within the recommended therapeutic reference interval. The groups of adherent and non-adherent patients did not differ statistically significantly in the severity of their illness as determined by the Clinical Global Impression (CGI), the Personal and Social Performance scale (PSP), and the Positive and Negative Syndrome Scale (PANSS). Conclusions: The combined method of evaluating adherence in schizophrenia patients confirmed the results determined by other methods. The benefits of this approach are described in the paper. © The Author(s) 2025.

关键词： Antipsychotics Schizophrenia Telehealth Therapeutic adherence Therapeutic drug monitoring

来源：评论

学校读者我要写书评

暂无评论

Content-Aware Opportunistic Packet Scheduling with Dynamic Routing Algorithm for Multi-path Video Streaming over Mesh Networks*

引用

Chinese Journal of Electronics 2023年第2期21卷 270-276页

作者： Yongfer Zhang Yunsheng Zhang Chun Yuan Shiyin Qin Zhihai He Digital Media Laboratory School of Computer Science and Engineering Beihang University Beijing China School of Automation Science and Electrical Engineering Beihang University Beijing China The Graduate School at Shenzhen Tsinghua University Shenzhen China Department of Electrical and Computer Engineering University of Missouri Columbia MO U.S.A

Multi-path video streaming over networks emerges as an important application in multimedia communication. Since the efficiency of multi-path video streaming is closely tied to the packet scheduling strategies, in this paper, we develop a content-aware opportunistic packet scheduling scheme with dynamic routing algorithm, which aims to maximize the reconstructed video quality by considering the content diversity of each video packet. To handle the problem of increased reordering delay, a sliding-window-based policy is then developed to find a balance between the reconstructed video quality and the reordering delay, throughout and packet loss. Our extensive experimental results demonstrate that the proposed packet scheduling scheme significantly outperforms EDPF and achieves a good tradeoff between the reconstructed video quality and the reordering time at the receiver side.

关键词： Scheduling algorithms Heuristic algorithms Packet loss Receivers Streaming media Routing Dynamic scheduling

来源：评论

学校读者我要写书评

暂无评论

Beyond Distribution Shift: Spurious Features Through the Lens of Training Dynamics

arXiv

引用

arXiv 2023年

作者： Murali, Nihal Puli, Aahlad Yu, Ke Ranganath, Rajesh Batmanghelich, Kayhan Intelligent Systems Program University of Pittsburgh United States Department of Computer Science New York University United States Department of Electrical and Computer Engineering Boston University United States

Deep Neural Networks (DNNs) are prone to learning spurious features that correlate with the label during training but are irrelevant to the learning problem. This hurts model generalization and poses problems when deploying them in safety-critical applications. This paper aims to better understand the effects of spurious features through the lens of the learning dynamics of the internal neurons during the training process. We make the following observations: (1) While previous works highlight the harmful effects of spurious features on the generalization ability of DNNs, we emphasize that not all spurious features are harmful. Spurious features can be "benign" or "harmful" depending on whether they are "harder" or "easier" to learn than the core features for a given model. This definition is model and dataset dependent. (2) We build upon this premise and use instance difficulty methods (like Prediction Depth (Baldock et al., 2021)) to quantify "easiness" for a given model and to identify this behavior during the training phase. (3) We empirically show that the harmful spurious features can be detected by observing the learning dynamics of the DNN’s early layers. In other words, easy features learned by the initial layers of a DNN early during the training can (potentially) hurt model generalization. We verify our claims on medical and vision datasets, both simulated and real, and justify the empirical success of our hypothesis by showing the theoretical connections between Prediction Depth and information-theoretic concepts like V-usable information (Ethayarajh et al., 2021). Lastly, our experiments show that monitoring only accuracy during training (as is common in machine learning pipelines) is insufficient to detect spurious features. We, therefore, highlight the need for monitoring early training dynamics using suitable instance difficulty metrics. © 2023, CC BY.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Human-Guided Artificial Intelligence (HGAI), a Framework for Overcoming AI's Blind Spots

Human-Guided Artificial Intelligence (HGAI), a Framework for...

引用

IEEE Southeastcon

作者： Ziyu Chen Selam B. Kelil Mohammad Y. M. Naser Jason S. Metcalfe Sylvia Bhattacharya Electrical & Computer Engineering Department Kennesaw State University Marietta USA Computer Science Department Kennesaw State University Marietta USA US DEVCOM Army Research Laboratory Aberdeen Proving Ground Aberdeen USA Engineering Technology Department Kennesaw State University Marietta USA

ISBN: (数字)9798331504847

ISBN: (纸本)9798331504854

Artificial Intelligence (AI) systems frequently exhibit systematic blind spots, often referred to as hallucinations in Large Language Models (LLMs), posing risks in high-stakes applications such as autonomous systems, security, and military operations. This paper explores how human intuitive responses, along with conscious reasoning processes, can be integrated to mitigate AI blind spots and enhance decision-making effectiveness within Human-AI teams. We introduce Human-Guided Artificial Intelligence (HGAI) as a framework for achieving this goal. Specifically, we examine the role of System-1 intuitive processing, as captured through physiological signals such as electroencephalography (EEG) and electrocardiography (ECG), System-2 reasoning-based decision-making, and Multi-Modal Fusion (MMF) mechanisms while assessing the knowledge required to develop more reliable, context-aware, and ethically aligned intelligent decision-making systems for highly complex environments.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Lightweight Wireless Sensing Through RIS and Inverse Semantic Communications

Lightweight Wireless Sensing Through RIS and Inverse Semanti...

引用

IEEE Conference on Wireless Communications and Networking

作者： Hongyang Du Jiacheng Wang Dusit Niyato Jiawen Kang Zehui Xiong Junshan Zhang Xuemin Sherman Shen The School of Computer Science and Engineering Nanyang Technological University Singapore The School of Automation Guangdong University of Technology China The Pillar of Information Systems Technology and Design Singapore University of Technology and Design Singapore The Department of Electrical and Computer Engineering University of California Davis USA The Department of Electrical and Computer Engineering University of Waterloo Canada

Thanks to the ubiquitous and easily accessible nature of wireless signals, wireless sensing is regarded as one of the promising techniques in the next-generation Internet of Things. In this paper, we propose the inverse semantic communications as a new paradigm to achieve lightweight wireless sensing using the reconfigurable intelligent surface (RIS). Instead of extracting semantic information from messages, we aim to encode the task-related source messages into a hyper-source message. Specifically, we first develop a novel RIS hardware for encoding several signal spectrums into one MetaSpectrum. We then propose a self-supervised learning method for decoding the MetaSpectrums to obtain the original signal spectrums. Using the sensing data collected from the real world, we show that our framework can reduce the data volume by 90% compared to that before encoding, without affecting the execution of various sensing tasks. Experiment results also demonstrate that the amplitude response matrix of the RIS enables the encryption of the sensing data.

关键词： Wireless communication Wireless sensor networks Semantics Self-supervised learning Encoding Sensors Decoding

来源：评论

学校读者我要写书评

暂无评论

NFTs for Trusted Traceability and Management of Digital Twins for Shipping Containers

NFTs for Trusted Traceability and Management of Digital Twin...

引用

IEEE Annual Conference on Pervasive Computing and Communications Workshops (PerCom)

作者： Feruz K. Elmay Mohammad Madine Khaled Salah Raja Jayaraman Department of Electrical Engineering and Computer Science Khalifa University Abu Dhabi UAE Department of Industrial and Systems Engineering Khalifa University Abu Dhabi UAE

The distribution, traceability, and management of shipping container logistics require secure data flow and trusted transactions. Digital Twins (DTs) can realize these features by offering shipping tracking and traceability, process flow and status monitoring, and management of the physical containersall in a remote manner. However, the data of a DT itself is typically stored, controlled, and managed by a centralized entity, which is often the original creator of the physical container. Having a centralized entity can cause mistrust. The centralized entity may alter, tamper, or delete the digital twin data. To overcome this problem, this paper proposes trusted sharing and management of DTs for shipping containers by using Non-Fungible Tokens (NFTs). NFTs are digital tokens that hold unique data stored, controlled, and managed in a decentralized and immutable blockchain ledger. We extend in this paper the use of NFTs to tokenize shipping container DTs and their metadata. The proposed solution uses NFTs and Ethereum blockchain smart contracts to offer decentralization, security, transparency, traceability, and immutability to the data and processes involved in the creation, storage, and management of DTs of shipping containers. To demonstrate our solution, we create a DT of a shipping container using Microsoft Azure Digital Twins services and showed how to tokenize it using NFT. We assess the system using various test cases to evaluate its main functionalities. Furthermore, we analyze the cost of transactions and the security of the smart contracts code. We have made the code of our smart contracts publicly available on GitHub.

关键词：

来源：评论

学校读者我要写书评

暂无评论

WARA-PS:a research arena for public safety demonstrations and autonomous collaborative rescue robotics experimentation

引用

Autonomous Intelligent systems 2021年第1期1卷 104-134页

作者： Olov Andersson Patrick Doherty Mårten Lager Jens-Olof Lindh Linnea Persson Elin A.Topp Jesper Tordenlid Bo Wahlberg Department of Computer and Information Science Linköping UniversityLinköpingSweden Saab Kockums MalmöSweden Division of Decision and Control Systems School of Electrical Engineering and Computer ScienceKTH Royal Institute of TechnologyStockholmSweden Department of Computer Science Lund UniversityLundSweden Saab Combitech AB LinköpingSweden Division of Automatic Control Department of Electrical EngineeringLinköping UniversityLinköpingSweden

A research arena(WARA-PS)for sensing,data fusion,user interaction,planning and control of collaborative autonomous aerial and surface vehicles in public safety applications is *** objective is to demonstrate scientific discoveries and to generate new directions for future research on autonomous systems for societal *** enabler is a computational infrastructure with a core system architecture for industrial and academic *** includes a control and command system together with a framework for planning and executing tasks for unmanned surface vehicles and aerial *** motivating application for the demonstration is marine search and rescue operations.A state-of-art delegation framework for the mission planning together with three specific applications is also *** first one concerns model predictive control for cooperative rendezvous of autonomous unmanned aerial and surface *** second project is about learning to make safe real-time decisions under uncertainty for autonomous vehicles,and the third one is on robust terrain-aided navigation through sensor fusion and virtual reality tele-operation to support a GPS-free positioning system in marine *** research results have been experimentally evaluated and demonstrated to industry and public sector audiences at a marine test *** would be most difficult to do experiments on this large scale without the WARA-PS research ***,these demonstrator activities have resulted in effective research dissemination with high public visibility,business impact and new research collaborations between academia and industry.

关键词： Autonomous systems Intelligent system architectures Research demonstration arena Autonomous drones Autonomous marine vessels Public safety and security Collaborative robotics

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：