检索结果-内蒙古大学图书馆

Proceedings of the Fortieth Conference on Uncertainty in Artificial Intelligence

作者： Jy-yong Sohn Dohyun Kwon Seoyeon An Kangwook Lee Department of Statistics and Data Science Yonsei University Republic of Korea Department of Mathematics University of Seoul Republic of Korea and Center for AI and Natural Sciences Korea Institute for Advanced Study Republic of Korea Department of Electrical and Computer Engineering University of Wisconsin-Madison WI

Fine-tuning large pre-trained models is a common practice in machine learning applications, yet its mathematical analysis remains largely unexplored. In this paper, we study fine-tuning through the lens of memorization capacity. Our new measure, the Fine-Tuning Capacity (FTC), is defined as the maximum number of samples a neural network can finetune, or equivalently, as the minimum number of neurons (m) needed to arbitrarily change N labels among K samples considered in the fine-tuning process. In essence, FTC extends the memorization capacity concept to the fine-tuning scenario. We analyze FTC for the additive fine-tuning scenario where the fine-tuned network is defined as the summation of the frozen pre-trained network f and a neural network g (with m neurons) designed for fine-tuning. When g is a ReLU network with either 2 or 3 layers, we obtain tight upper and lower bounds on FTC; we show that N samples can be fine-tuned with m = Θ(N) neurons for 2-layer networks, and with m = Θ(√N) neurons for 3-layer networks, no matter how large K is. Our results recover the known memorization capacity results when N = K as a special case.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Femtosecond Laser based Post-Processing Free True Random Number Generator

Femtosecond Laser based Post-Processing Free True Random Num...

引用

Frontiers in Optics + Laser Science 2023, FiO, LS 203: Part of Frontiers in Optics + Laser Science 2023

作者： Sharma, Amritash Paithankar, Prasanna Varshney, Shailendra K. Group Advanced Photonics Laboratory Department of Electronics and Electrical Communication Engineering Indian Institute of Technology West Bengal Kharagpur721302 India Department of Applied Optics and Photonics University of Calcutta Technology Campus JD 2 Salt Lake Kolkata700098 India Department of Computer Science and Engineering Indian Institute of Technology West Bengal Kharagpur721302 India Authors contributed equally

We experimentally demonstrate a high-entropy source based on the intensity fluctuation in a femtosecond laser. The setup successfully passes statistical tests without any post-processing. The generator can be potentia... 详细信息

ISBN: (纸本)9781957171296

关键词： Femtosecond lasers Laser sources Mode-locked fiber lasers Quantum communications Quantum cryptography Ultrafast lasers

来源：评论

学校读者我要写书评

暂无评论

Photonic Bragg gratings with cladding asymmetry for polarization independent and rotation filter

Photonic Bragg gratings with cladding asymmetry for polariza...

引用

2023 Conference on Lasers and Electro-Optics, CLEO 2023

作者： Pimbi, Daniel Hasan, Mehedi Mia, Md Borhan Jaidye, Nafiz Kim, Sangsik Department of Electrical and Computer Engineering Texas Tech University LubbockTX79409 United States Nano Tech Center Texas Tech University LubbockTX79409 United States Department of Physics and Astronomy Texas Tech University LubbockTX79409 United States School of Electrical Engineering Korea Advanced Institute of Science and Technology Daejeon34141 Korea Republic of

We present a single-etched polarization-independent Bragg filter using cladding asymmetry. An air cladding strongly couples TE and TM modes, rotating its polarization status and achieving polarization - in sensitive B... 详细信息

ISBN: (纸本)9781957171258

关键词： Polarization

来源：评论

学校读者我要写书评

暂无评论

Machine Learning and IoT based Performance Improvement of Energy Efficiency in Smart Buildings 2

Machine Learning and IoT based Performance Improvement of En...

引用

2nd International Conference on Sustainable Computing and Data Communication Systems, ICSCDS 2023

作者： Jagadeesan, Sowmya Ravi, C.N. Sujatha, M. Southry, S. Sree Sundararajan, J. Reddy, Ch. Venkata Krishna Vels Institute of Science Technology and Advanced Studies Departnent of Bca and It INurture Education Solutions Tamil Nadu Chennai India Cmr Engineering College Department of Computer Science and Engineering Telangana Hyderabad India Easwari Engineering College Ramapuram Department of Electrical and Electronics Engineering Tamil Nadu Chennai India Sona College of Technology Department of Electronics and Communication Engineering Tamil Nadu Salem India Npr College of Enggineering and Technology Department of Electronics and Communication Tamil Nadu Dindugul India Chaitanya Bharathi Institute of Technology Department of Electrical and Electronics Engineering Telangana Hyderabad India

ISBN: (纸本)9781665491990

The energy plays a prominent role in the development of a country's economic conditions. This helps to increase the overall performance index of a country. Thus a continuous energy system is needed to develop a country's performance index. Thus to have an advancement in the energy management system, an improved system called smart grid is implemented. This helps in efficient usage of energy management through meeting the demand side managing in the smart buildings. This assistances to decrease the additional procedure of energy consumption through optimization techniques such as genetic algorithm used in smart buildings. This is done with the machine learning through optimization algorithm. This helps to proper scheduling of appliances. This is done through the priority scheduling. The overall performance is done through the internet of things with sensors. The sensors helps in the process of sensing the external physical parameters such as temperature, humidity, moisture, heat and light. This helps to enhance the energy efficiency in the smart buildings through the machine learning and IoT. © 2023 IEEE.

关键词： Internet of things

来源：评论

学校读者我要写书评

暂无评论

Biologically-Informed Excitatory and Inhibitory Balance for Robust Spiking Neural Network Training

arXiv

引用

arXiv 2024年

作者： Kilgore, Joseph A. Kopsick, Jeffrey D. Ascoli, Giorgio A. Adam, Gina C. Department of Electrical and Computer Engineering George Washington University WashingtonDC20052 United States Center for Neural Informatics Structures & Plasticity Krasnow Institute for Advanced Study George Mason University FairfaxVA22030 United States

Spiking neural networks drawing inspiration from biological constraints of the brain promise an energy-efficient paradigm for artificial intelligence. However, challenges exist in identifying guiding principles to train these networks in a robust fashion. In addition, training becomes an even more difficult problem when incorporating biological constraints of excitatory and inhibitory connections. In this work, we identify several key factors, such as low initial firing rates and diverse inhibitory spiking patterns, that determine the overall ability to train spiking networks with various ratios of excitatory to inhibitory neurons on AI-relevant datasets. The results indicate networks with the biologically realistic 80:20 excitatory:inhibitory balance can reliably train at low activity levels and in noisy environments. Additionally, the Van Rossum distance, a measure of spike train synchrony, provides insight into the importance of inhibitory neurons to increase network robustness to noise. This work supports further biologically-informed large-scale networks and energy efficient hardware implementations. Copyright © 2024, The Authors. All rights reserved.

关键词： Neural networks

来源：评论

学校读者我要写书评

暂无评论

Dual-Port Endfire Millimeter Wave Reconfigurable Antenna with Optimized Pixel Surface

Dual-Port Endfire Millimeter Wave Reconfigurable Antenna wit...

引用

Antennas and Propagation Society International Symposium

作者： Shiwen Tang Yujie Zhang Junhui Rao Tianrui Qiao Chi-Yuk Chiu Ross Murch Department of Electronic and Computer Engineering Institute for Advanced Study The Hong Kong University of Science and Technology Clear Water Bay Kowloon Hong Kong

A dual-port 1×2 antenna array incorporating a parasitic pixel surface in front of the array is proposed. By optimizing the pixel surface, different radiation patterns can be provided by each port. To demonstrate the effectiveness of the approach, a dual-port design, with patterns directed to 112° and 66° and gains of 7.87 dBi and 6.95 dBi respectively at 28 GHz, is provided. A feature of the design is that no RF switches are needed in the pixel surface. The proposed antenna array can be suitable for future fifth-generation (5G) wireless communication systems.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Blockchain-Enabled Federated Learning with Neuromorphic Edge Devices for Drone Identification and Flight Mode Detection

Blockchain-Enabled Federated Learning with Neuromorphic Edge...

引用

Conference on Digital Avionics Systems (DASC)

作者： Alex Henderson Chris Yakopcic Jamison Colter Steven Harbour Tarek Taha Dayton Engineering Advanced Projects Lab Southwest Research Institute Beavercreek OH Department of Electrical and Computer Engineering University of Dayton Dayton OH

Unmanned aerial vehicles (UAVs), also known as drones, are expected to play an integral role in next-generation wireless avionics networks. To help secure these networks from malicious activity, machine learning (ML) based approaches have been proposed to identify drones and their flight modes via direct radio frequency signals. These approaches frequently rely on cloudcentric methods for training, which present serious concerns, such as privacy leakage, resource burden, and undesirable latency. In response to these concerns, a distributed ML paradigm known as federated learning (FL) has been proposed that enables multiple drones to collaboratively train ML models by only exchanging model parameters and not the raw data itself. Unfortunately, the conventional FL framework is strongly dependent on a centralized aggregation server with multiple resource-constrained edge devices, making it susceptible to poisoning attacks and poor network utilization. Recently, it has been found that blockchain technology (BT) holds significant promise for securing and storing data in edge applications with high levels of trust. However, the expanding chain of security blocks consumes a significant amount of computing power, thus limiting its scalability. To bridge the gap between computational efficiency and security, we propose a blockchain-empowered and energy-efficient FL framework with neuromorphic edge devices. To assess the effectiveness of our proposed framework we perform a drone identification and flight mode detection task with a spiking neural network (SNN). Finally, we compare our neuromorphic approach with competitive alternatives to validate the energy and performance gains of our system.

关键词：

来源：评论

学校读者我要写书评

暂无评论

GANscan:continuous scanning microscopy using deep learning deblurring

引用

Light(Science & Applications) 2022年第10期11卷 2378-2387页

作者： Michael John Fanous Gabriel Popescu Quantitative Light Imaging Laboratory Beckman Institute for Advanced Science and TechnologyUniversity of Illinois at Urbana-ChampaignUrbanaIL 61801USA Department of Bioengineering University of Illinois at Urbana-Champaign306 N.Wright StreetUrbanaIL 61801USA Department of Electrical and Computer Engineering University of Illinois at Urbana-Champaign306 N.Wright StreetUrbanaIL 61801USA

Most whole slide imaging(WSI)systems today rely on the"stop-and-stare"approach,where,at each field of view,the scanning stage is brought to a complete stop before the camera snaps a *** procedure ensures that each image is free of motion blur,which comes at the expense of long acquisition *** order to speed up the acquisition process,especially for large scanning areas,such as pathology slides,we developed an acquisition method in which the data is acquired continuously while the stage is moving at high *** generative adversarial networks(GANs),we demonstrate this ultra-fast imaging approach,referred to as GANscan,which restores sharp images from motion blurred *** allows us to complete image acquisitions at 30x the throughput of stop-and-stare *** method is implemented on a Zeiss Axio Observer Z1 microscope,requires no specialized hardware,and accomplishes successful reconstructions at stage speeds of up to 5000 μm/*** validate the proposed method by imaging H&E stained tissue *** method not only retrieves crisp images from fast,continuous scans,but also adjusts for defocusing that occurs during scanning within+/-5 μ*** a consumer GPU,the inference runs at<20 ms/image.

关键词： hardware continuous image

来源：评论

学校读者我要写书评

暂无评论

Language models as science tutors 24

Language models as science tutors

引用

Proceedings of the 41st International Conference on Machine Learning

作者： Alexis Chevalier Jiayi Geng Alexander Wettig Howard Chen Sebastian Mizera Toni Annala Max Jameson Aragon Arturo Rodríguez Fanlo Simon Frieder Simon Machado Akshara Prabhakar Ellie Thieu Jiachen T. Wang Zirui Wang Xindi Wu Mengzhou Xia Wenhan Xia Jiatong Yu Jun-Jie Zhu Zhiyong Jason Ren Sanjeev Arora Danqi Chen Princeton Language and Intelligence Princeton University School of Natural Sciences Institute for Advanced Study School of Mathematics Institute for Advanced Study Neuroscience Institute Princeton University Hebrew University of Jerusalem Department of Computer Science Oxford University FIM - Institute for Mathematical Research ETH Zürich Department of Computer Science University of Wisconsin-Madison and Meta FAIR Department of Civil and Environmental Engineering and Andlinger Center for Energy and the Environment Princeton University

NLP has recently made exciting progress toward training language models (LMs) with strong scientific problem-solving skills. However, model development has not focused on real-life use-cases of LMs for science, including applications in education that require processing long scientific documents. To address this, we introduce TUTOREVAL and TUTORCHAT. TUTOREVAL is a diverse question-answering benchmark consisting of questions about long chapters from STEM textbooks, written by experts. TUTOREVAL helps measure real-life usability of LMs as scientific assistants, and it is the first benchmark combining long contexts, free-form generation, and multi-disciplinary scientific knowledge. Moreover, we show that fine-tuning base models with existing dialogue datasets leads to poor performance on TUTOREVAL. Therefore, we create TUTORCHAT, a dataset of 80,000 long synthetic dialogues about textbooks. We use TUTORCHAT to fine-tune Llemma models with 7B and 34B parameters. These LM tutors specialized in math have a 32K-token context window, and they excel at TUTOREVAL while performing strongly on GSM8K and MATH. Our datasets build on open-source materials, and we release our models, data, and evaluations publicly.

关键词：

来源：评论

学校读者我要写书评

暂无评论

IoT Farm: A Robust Methodology Design to Support Smart Agricultural System Using Internet of Things with Intelligent Sensors Association 7

IoT Farm: A Robust Methodology Design to Support Smart Agric...

引用

7th International Conference on Electronics, Communication and Aerospace Technology, ICECA 2023

作者： Karthiga, R. Brindha Devi, C.L. Janaki, R. Gayathri, C. Samuthira Pandi, S. Shobana, D. R.M.K Engineering College Kavaraipettai Department of IT Tamilnadu Chennai India Department of Computer Science Tamilnadu Chennai India Prince Shri Venkateshwara Padmavathy Engineering College Department of Science & Humanities Tamilnadu Chennai India Chennai Institute of Technology Centre for Advanced Wireless Integrated Technology Tamilnadu Chennai India Rajalakshmi Engineering College Department of Electrical and Electronics Engineering Tamilnadu Chennai India

ISBN: (纸本)9798350340600

The term 'Smart Farming' has been increasingly popular in recent years due, in large part, to the proliferation of network-based, high-tech farm monitoring cycles. These cycles make extensive use of technologies for communication and information gathering in the form of machinery, equipment, and sensors. The Internet of Things, often known as IoT, is the wave of the future in every sector of the economy, and its impacts will be felt by everyone. It is a collection of different devices that are capable of configuring themselves into a fully operational network. Conventional farming practices are progressively being supplanted by smart farming, which takes use of advancements made possible by the internet of things. This makes conventional farming methods more efficient, which in turn saves farmers money and reduces the amount of food that is lost. This technology helps farmers conserve energy and time by evaluating data in actual time from sensors placed in crops and offering actionable information on how their crops are progressing. The data comes from the sensors themselves, which are planted in the crops. Data from the field is sent to the cloud and evaluated there so that procedures may be automated utilizing Internet of Things devices. The assessment of the farmer's contribution into the process of making choices is demonstrated with information regarding the temperature of the field, the soil moisture level, the humidity level, and the amount of water that is used. In addition, the technical knowledge presented here is valuable to producers at every stage of the production process, including planting, growing, harvesting, packaging and shipping. © 2023 IEEE.

关键词： Internet of things

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：