检索结果-内蒙古大学图书馆

61st ACM/IEEE Design Automation Conference, DAC 2024

作者： Shin, Dongjin Choi, Insu Yang, Joon-Sung Department of Electrical and Electronic Engineering Yonsei University Seoul Korea Republic of Department of Systems Semiconductor Engineering Yonsei University Seoul Korea Republic of BK21 Graduate Program in Intelligent Semiconductor Technology Yonsei University Seoul Korea Republic of

ISBN: (纸本)9798400706011

Vision Transformers have demonstrated remarkable performance in various vision tasks. However, general-purpose processors, such as CPUs and GPUs, face challenges in efficiently handling the inference of Vision Transformers. To address the issue, prior works have focused on accelerating only attention due to its high computational cost in NLP Transformers. In contrast, Vision Transformers demonstrate a higher computational cost due to linear modules such as linear transformation, linear projection and Feed-Forward Network (FFN), compared to attention. In this paper, we present ViT-slice, an algorithm-architecture co-design that enhances end-to-end performance and energy efficiency by optimizing not only attention but also linear modules. At the algorithm level, we propose bit-slice compression that avoids storing the redundant most significant bits (MSBs). Additionally, we present bit-slice dot product with early skip to efficiently compute the dot product using bit-sliced data. To enable early skip during the dot product computation, we leverage a trainable threshold. On the hardware level, we introduce a specialized bit-slice dot product unit (BSDPU) to efficiently process the bit-slice dot product with early skip algorithm. Additionally, we present a bit-slice encoder and decoder for on-chip bit-slice compression. ViT-slice achieves 244×, 35.3×, 16.8×, 10.4×, 5.0× end-to-end speedup over Xeon CPU, EdgeGPU, TITAN Xp GPU, Sanger accelerator and ViTCoD accelerator, respectively. © 2024 Copyright held by the owner/author(s).

关键词： Linear transformations

来源：评论

学校读者我要写书评

暂无评论

Text-to-Speech User Interface for ChatGPT 4

Text-to-Speech User Interface for ChatGPT

引用

4th International Conference on Electronic and electrical engineering and Intelligent System, ICE3IS 2024

作者： Suni, Alfa Faridh Utomo, Aryo Baskoro Fathoni, Khoiruddin Mahendra, Budiandra Yusuf Seinsiani, Izzati Gemi Hastawan, Ahmad Fashiha Universitas Negeri Semarang Computer Engineering Semarang Indonesia Universitas Negeri Semarang Electrical Engineering Semarang Indonesia Universitas Negeri Semarang Information Technology Education Semarang Indonesia Universitas Negeri Semarang English Literature Study Program Semarang Indonesia

ISBN: (纸本)9798350378368

The emergence of OpenAI's ChatGPT is paying a lot of attention to Generative AI and its impact on Academic integrity. Generative AI is a system designed to generate content or output such as text, image, audio, simulation, video and code based on their respective training data. ChatGPT itself is in the form of a chatbot where users can communicate with AI through dialogue applications via text. GPT 3 API is the technology behind the ChatGPT. However, interaction via text is not as flexible as dialogue using voice. Therefore, speech-to-text technology is needed to interface between the user and the AI. Chatbot AI powered by GPT3 with the optional voice UI powered by Google Text to Speech clearly benefits accessibility and usability. The output of dialogue between human users and AI can be done through text or sound. This research attempts to implement a Text to Speech and vice versa Speech to Text as a User Interface on a Gen-AI using the GPT 3 API so that it will be more user-friendly and accessible. This research opens many possibilities for human-AI interaction (HAX) in implementing virtual assistants. A seamless dialogue between humans and AI will be our goal in the future. © 2024 IEEE.

关键词： ChatGPT Generative AI Speech to Text Text to Speech

来源：评论

学校读者我要写书评

暂无评论

Transfer Learning Evaluation of Convolutional Neural Network Architectures for Thermal Image Classification of Solar Modules

Transfer Learning Evaluation of Convolutional Neural Network...

引用

International electrical engineering Congress (iEECON)

作者： Chutinun Phungket Prasit Nangtin The Graduate Program Faculty of Engineering Faculty of Engineering Pathumwan Institute of Technology Bangkok Thailand Department of Electrical Engineering Faculty of Engineering Pathumwan Institute of Technology Bangkok Thailand

ISBN: (数字)9798331543952

ISBN: (纸本)9798331543969

Accurate anomaly classification in solar modules is crucial for efficient solar power system monitoring and maintenance. This study explores deep learning, specifically Convolutional Neural Networks (CNNs), for automated defect classification in thermal images. We employed transfer learning with pre-trained AlexNet and GoogleNet architectures, comparing static feature extraction and fine-tuning. Using a dataset of thermal images categorized as no anomaly, cracked, or soiled modules, the fully fine-tuned GoogleNet model achieved the highest classification accuracy 95.5%, outperforming AlexNet and traditional feature-based methods. GoogleNet's superior performance stems from its Inception modules, which learn more discriminative features by employing multiple convolutional filter sizes within the same layer, capturing complex thermal image patterns. Fine-tuning further boosts performance by adapting pre-trained knowledge to the specific task. These results demonstrate deep learning's potential for enhancing solar system maintenance reliability and efficiency.

关键词： Deep learning Adaptation models Accuracy Transfer learning Solar energy Feature extraction Maintenance Solar panels Convolutional neural networks Image classification

来源：评论

学校读者我要写书评

暂无评论

Ultra-Thin CMOS-Compatible Artificial Magnetic Conductor for Millimeter-Wave Antenna-on-Chip

Ultra-Thin CMOS-Compatible Artificial Magnetic Conductor for...

引用

Antennas and Propagation & USNC/URSI National Radio Science Meeting

作者： Yiyang Yu Atif Shamim Electrical and Computer Engineering Program King Abdullah University of Science and Technology Thuwal KSA

ISBN: (数字)9798350369908

ISBN: (纸本)9798350369915

Antenna-on-chip (AoC) is one of the promising solutions in the 6G era. However, it suffers from poor radiation due to the lossy Si substrate. Placing an artificial magnetic conductor (AMC) under the AoC cannot only isolate the AoC from the lossy Si substrate but also provide in-phase reflection. Nonetheless, the ultra-thin stack-up of the mainstream semiconductor technology makes the realization of AMC challenging at the millimeter-wave spectrum. To make the AMC electrically thicker, this paper presents four kinds of thickness reduction techniques (TRT): slot, notch, metallic post, and embedded guiding structure. With TRTs, an ultra-thin AMC ( $15\ \mu \mathrm{m}$ thickness) is realized. With the proposed AMC, the AoC demonstrated a gain of 3.2 dB at 72 GHz. Compared to the standalone antenna, the AMC enhances the gain by 7 dB.

关键词： 6G mobile communication Conferences Millimeter wave technology Conductors Silicon Reflection Gain

来源：评论

学校读者我要写书评

暂无评论

Compact Refractive Error Measuring Device 6

Compact Refractive Error Measuring Device

引用

6th IEEE Eurasian Conference on Educational Innovation, ECEI 2023

作者： Yeh, Feng-Ming Chiang, Chun-Yu Chang, Chao-Kai Wang, Hsuan-Fu Tien, Chuen-Lin Chen, Der-Chin Department of Optometry Yuanpei University of Medical Technology Hsinchu Taiwan Ph.D. Program of Electrical and Communications Engineering Feng Chia University Taichung Taiwan Department of Aeronautical Engineering National Formosa University Yunlin Taiwan Department of Electrical Engineering Feng Chia University Taichung Taiwan

ISBN: (纸本)9798350334289

In this study, a reflective aspheric optical system is used to establish an eye examination device for distance and near vision tests. The system includes a micro-display with a rotating device, an off-axis parabolic mirror, an off-axis ellipsoid mirror, a chin-fixing bracket, and vision software. The display contents of the micro-display include English C and E charts, Chinese words, graphics, and animation. The off-axis parabolic mirror and off-axis ellipsoid mirror are confocal, and the micro-display is placed at the confocal point. The micro-display is projected through an off-axis parabolic mirror for myopia's distance vision testing, and an off-axis elliptical mirror for myopia's near vision testing. Others, the graphics and animation, provide visual checks for young children. In addition, CCD cameras and AI recognition can be used as visual examinations for people with language impairment. In this study, twenty-five subjects with the refractive error were tested using E-chart and this device to obtain the decimal point VA value, and the trial frame and lens were used to obtain the diopter of glasses. The date of the experimental results was used as a reference value for ophthalmological examination. This system uses an off-axis parabolic mirror and an off-axis ellipsoidal mirror. Optical alignment is a challenge to solve for the system. Students participating in the study obtained good optics training. The advantages of the developed device include the convenience to carry with a multi-purpose and straightforward use. © 2023 IEEE.

关键词： Mirrors

来源：评论

学校读者我要写书评

暂无评论

Super-Resolution Displacement Spectroscopic Sensing over a Surface“Rainbow”

引用

engineering 2022年第10期17卷 75-81页

作者： Lyu Zhou Nan Zhang Chang Chieh Hsu Matthew Singer Xie Zeng Yizheng Li Haomin Song Josep Jornet Yun Wu Qiaoqiang Gan Department of Electrical Engineering The State University of New York at BuffaloBuffaloNY 14260USA Department of Biomedical Engineering The State University of New York at BuffaloBuffaloNY 14260USA Department of Electrical and Computer Engineering Northeastern UniversityBostonMA 02115USA Material Science Engineering Program King Abdullah University of Science and TechnologyThuwal 23955-6900Saudi Arabia

Subwavelength manipulation of light waves with high precision can enable new and exciting applications in spectroscopy,sensing,and medical *** these applications,miniaturized spectrometers are desirable to enable the on-chip analysis of spectral *** particular,for imaging-based spectroscopic sensing mechanisms,the key challenge is to determine the spatial-shift information accurately(i.e.,the spatial displacement introduced by wavelength shift or biological or chemical surface binding),which is similar to the challenge presented by super-resolution ***,we report a unique"rainbow"trapping metasurface for on-chip spectrometers and *** with super-resolution image processing,the low-setting 4×optical microscope system resolves a displacement of the resonant position within 35 nm on the plasmonic rainbow trapping metasurface with a tiny area as small as0.002 *** unique feature of the spatial manipulation of efficiently coupled rainbow plasmonic resonances reveals a new platform for miniaturized on-chip spectroscopic analysis with a spectral resolution of 0.032 nm in wavelength *** this low-setting 4×microscope imaging system,we demonstrate a biosensing resolution of 1.92×109exosomes per milliliter for A549-derived exosomes and distinguish between patient samples and healthy controls using exosomal epidermal growth factor receptor(EGFR)expression values,thereby demonstrating a new on-chip sensing system for personalized accurate bio/chemical sensing applications.

关键词： Rainbow trapping Metasurface Surface plasmon polaritons Super-resolution displacement On-chip biosensing

来源：评论

学校读者我要写书评

暂无评论

Reachability Analysis of Nonlinear Discrete-Time Systems Using Polyhedral Relaxations and Constrained Zonotopes 63

Reachability Analysis of Nonlinear Discrete-Time Systems Usi...

引用

63rd IEEE Conference on Decision and Control, CDC 2024

作者： Rego, Brenner S. Raffo, Guilherme V. Terra, Marco H. Scott, Joseph K. Georgia Institute of Technology Department of Chemical and Biomolecular Engineering 311 Ferst Dr. AtlantaGA30318 United States University of Säo Paulo Department of Electrical and Computer Engineering SP Säo Carlos13566-590 Brazil Federal University of Minas Gerais Department of Electronics Engineering and the Graduate Program in Electrical Engineering MG Belo Horizonte31270-901 Brazil

ISBN: (纸本)9798350316339

This paper presents a novel algorithm for reachability analysis of nonlinear discrete-time systems. The proposed method combines constrained zonotopes (CZs) with polyhedral relaxations of factorable representations of nonlinear functions to propagate CZs through nonlinear functions, which is normally done using conservative linearization techniques. The new propagation method provides better approximations than those resulting from linearization procedures, leading to significant improvements in the computation of reachable sets in comparison to other CZ methods from the literature. Numerical examples highlight the advantages of the proposed algorithm. © 2024 IEEE.

关键词： Numerical methods

来源：评论

学校读者我要写书评

暂无评论

Multiplexed Nanoscopy via Buffer Exchange

引用

ACS Nano 2024年第34期18卷 23445-23456页

作者： Chang, Ting-Jui Ben Yang, T. Tony Department of Electrical Engineering National Taiwan University Taipei10617 Taiwan Graduate Institute of Biomedical Electronics and Bioinformatics National Taiwan University Taipei10617 Taiwan Department of Physics National Taiwan University Taipei10617 Taiwan Nano Science and Technology Program Taiwan International Graduate Program Academia Sinica and National Taiwan University Taipei10617 Taiwan

Understanding cellular functions, particularly in their intricate complexity, can greatly benefit from the spatial mapping of diverse molecules through multitarget single-molecule localization microscopy (SMLM). Existing methodologies, primarily restricting the encoding dimensions to color and lifetime or requiring cyclic staining, often involve broad chromatic detection, specialized optical configurations, or sophisticated labeling techniques. Here, we propose a simple approach called buffer-exchange stochastic optical reconstruction microscopy (beSTORM), which introduces an additional dimension to differentiate between single molecules irrespective of their spectral properties. This method leverages the distinguishable photoblinking responses to distinct buffer conditions, offering a straightforward yet effective means of fluorophore discrimination. Through buffer exchanges, beSTORM achieves multitarget SMLM imaging with minimal crosstalk. Direct integration with expansion microscopy (ExM) demonstrates its capability to resolve up to six proteins at the molecular level within a single emission color without chromatic aberration. Overall, beSTORM presents a highly compatible imaging platform, promising significant advancements in highly multiplexed nanoscopy for exploring multiple targets in biological systems with nanoscale precision. © 2024 The Authors. Published by American Chemical Society.

关键词： Stochastic systems

来源：评论

学校读者我要写书评

暂无评论

Deep-Learning-Based Magnetic Resonance Simultaneous Multislice Imaging Using Holographic Image Decoding

Deep-Learning-Based Magnetic Resonance Simultaneous Multisli...

引用

IEEE International Conference on Image Processing

作者： Satoshi Ito Yuki Sato Naoya Endo Shohei Ouchi Electrical and Electronic Systems Engineering Graduate Program in Information Utsunomiya University Department of Innovative Electrical and Electronic Engineering National Institute of Technology Oyama

ISBN: (数字)9798350349399

ISBN: (纸本)9798350349405

We propose a new magnetic resonance simultaneous multislice imaging (SMS) by developing our previously proposed imaging method based on the principle of holography. Different amplitude modulations for each slice are applied to separate the slice images. The slice image separation is performed using a data-driven science in which U-Nets separates the slice images using the given amplitude modulation as a cue. Simulation experiments using a real-valued and an phase-varied image model revealed the relationship between the coefficients of amplitude modulation, the number of simultaneously excited slices, and image quality. In addition, we performed image separation experiments on the holographic imaging method by approximating the excited volume region as five slice images with no gaps, and showed the feasibility of image separation. This method has the potential to be a new type of SMS.

关键词： Image quality Deep learning Phase modulation Neural networks Magnetic resonance Holography Amplitude modulation

来源：评论

学校读者我要写书评

暂无评论

Interference Management Using Power Allocation for Downlink Transmission of Femtocell-Macrocell Ultra Dense Networks 9

Interference Management Using Power Allocation for Downlink ...

引用

9th International Conference on Information technology, Computer and electrical engineering, ICITACEE 2022

作者： Suryani, Sri Susanto, Misfa Sadnowo, Ageng Hasim, Sitronella Nurfitriani Hamdani, Fadil University of Lampung Bandar Department of Electrical Engineering Lampung Indonesia Institute Technology of Sumatera South Study Program of Telecommunication Engineering Lampung Indonesia

ISBN: (数字)9781665471503

ISBN: (纸本)9781665471480

Femtocell is a promising technology to increase network capacity and coverage, especially in indoor areas. It has low power, short range. The Home Base Station/HeNB (Femtocell Base Station) can be installed by users using connection of personal equipment such as laptops, notepads and cell phones, and it is activated by service provider. By deploying femtocells into existing macrocell/microcell areas, the radio resources are shared among femtocells and existing macrocells. However, the sharing of limited available bandwidth frequency licenses cause interference problems between femtocells and/or macrocells. The interference problem becomes more complex when the number of deployed femtocells is many and dense creating a phenomenon of Ultra Dense Network (UDN). To cope with this interference problem, this paper proposes a power allocation method for femtocells deployed in macrocells areas (UDN) at downlink transmissions. In this method the power transmitted at HeNB to each femtocell user is adjusted in such a way as to reduce or increase the transmitted power at HeNB. If the estimated SINR value at current time is less than 20 dB, the power on the HeNB is increased by such a way so SINR is expectedly to be 20 dB for next transmission time. However, if the obtained SINR at current transmission time is more than 20 dB, then the transmission power at the HeNB is reduced so that the SINR value becomes 20 dB for the next transmission time. Extensive simulation has been carried out and the performance parameters in terms of SINR, throughput, and BER have been measured and compared to the baseline system, i.e., the system without power allocation method. The simulation results show that the system that applies power allocation method which is used in the cell edge area is better than the baseline system both in the center and the edge of cells. © 2022 IEEE.

关键词： Femtocell

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：