检索结果-内蒙古大学图书馆

2024 Conference on Empirical Methods in Natural Language Processing, EMNLP 2024

作者： Kim, Eunji Shim, Kyuhong Chang, Simyung Yoon, Sungroh Department of Electrical and Computer Engineering Seoul National University Korea Republic of Qualcomm AI Research an initiative of Qualcomm Technologies Inc. Qualcomm Korea YH Korea Republic of Interdisciplinary Program in Artificial Intelligence Seoul National University Korea Republic of

ISBN: (纸本)9798891761681

A text encoder within Vision-Language Models (VLMs) like CLIP plays a crucial role in translating textual input into an embedding space shared with images, thereby facilitating the interpretative analysis of vision tasks through natural language. Despite the varying significance of different textual elements within a sentence depending on the context, efforts to account for variation of importance in constructing text embeddings have been lacking. We propose a framework of Semantic Token Reweighting to build Interpretable text embeddings (SToRI), which incorporates controllability as well. SToRI refines the text encoding process in CLIP by differentially weighting semantic elements based on contextual importance, enabling finer control over emphasis responsive to data-driven insights and user preferences. The efficacy of SToRI is demonstrated through comprehensive experiments on few-shot image classification and image retrieval tailored to user preferences. © 2024 Association for Computational Linguistics.

关键词： Embeddings

来源：评论

学校读者我要写书评

暂无评论

Enhancing the Performance of E-Mode AlGaN/GaN HEMTs With Recessed Gates Through Low-Damage Neutral Beam Etching and Post-Metallization Annealing

IEEE Open Journal of Nanotechnology

引用

IEEE Open Journal of Nanotechnology 2023年 4卷 150-155页

作者： Chen, Yi-Ho Ohori, Daisuke Aslam, Muhammad Lee, Yao-Jen Li, Yiming Samukawa, Seiji National Yang Ming Chiao Tung University Parallel and Scientific Computing Laboratory Graduate Degree Program of College of Electrical and Computer Engineering Hsinchu300093 Taiwan Tohoku University Institute of Fluid Science Sendai980-8577 Japan National Yang Ming Chiao Tung University Parallel and Scientific Computing Laboratory Electrical Engineering and Computer Science International Graduate Program Hsinchu300093 Taiwan National Yang Ming Chiao Tung University Institute of Pioneer Semiconductor Innovation Hsinchu300093 Taiwan National Yang Ming Chiao Tung University Parallel and Scientific Computing Laboratory Institute of Communications Engineering Institute of Biomedical Engineering Hsinchu300093 Taiwan National Yang Ming Chiao Tung University Department of Electronics and Electrical Engineering Hsinchu300093 Taiwan National Yang Ming Chiao Tung University Institute of Communications Engineering Hsinchu300093 Taiwan

This study investigated the electrical properties of AlGaN/GaN high-electron-mobility transistors (HEMTs) with varied recess depths under the gate electrode. We demonstrated a recess depth of approximately 6 nm, which was achieved through neutral beam etching (NBE) technique with a low etch rate of 1.8 nm/min, resulting in device enhancement-mode (E-mode) behavior with threshold voltage (Vth) of 0.49 V. The effects of post-metallization annealing (PMA) on the device performance were also examined. The results revealed that PMA treatment improves the DC characteristics of the devices, including maximum drain current (IDMAX), transconductance (gm), subthreshold swing (SS), on-off ratio, and off-state leakage current, with maximum enhancement percentage of 18.3% for IDMAX, 3758% for on-off ratio, and 54.3% for SS. Moreover, this study compared the recess depths of metal-insulator-semiconductor high-electron-mobility transistors (MIS-HEMTs) with the SiN dielectric layer. The results showed that MIS-HEMTs exhibit more negative Vth values, which can be attributed to the controlled surface states achieved through passivation. © 2020 IEEE.

关键词： Etching

来源：评论

学校读者我要写书评

暂无评论

Machine-Learning-Enhanced Quantum Optical Storage in Solids

arXiv

引用

arXiv 2024年

作者： Lei, Yisheng An, Haechan Li, Zongfeng Hosseini, Mahdi Department of Electrical and Computer Engineering Applied Physics Program Northwestern University EvanstonIL60208 United States Elmore Family School of Electrical and Computer Engineering Purdue University West LafayetteIN47907 United States

Quantum memory devices with high storage efficiency and bandwidth are essential elements for future quantum networks. Solid-state quantum memories can provide broadband storage, but they primarily suffer from low storage efficiency. We use passive optimization and machine learning techniques to demonstrate nearly a 6-fold enhancement in quantum memory efficiency. In this regime, we demonstrate coherent and single-photon-level storage with a high signal-to-noise ratio. The optimization technique presented here can be applied to most solid-state quantum memories to significantly improve the storage efficiency without compromising the memory bandwidth. © 2024, CC BY.

关键词： Signal to noise ratio

来源：评论

学校读者我要写书评

暂无评论

Efficient Pumping of Spectral Holes in a Tm3+: YAG Crystal for Broadband Quantum Optical Storage

arXiv

引用

arXiv 2024年

作者： Lei, Yisheng Li, Zongfeng Hosseini, Mahdi Department of Electrical and Computer Engineering and Applied Physics Program Northwestern University EvanstonIL60208 United States Elmore Family School of Electrical and Computer Engineering Purdue University West LafayetteIN47907 United States

Quantum memory devices with high storage efficiency and bandwidth are essential elements for future quantum networks. Here, we report a storage efficiency greater than 28% in a Tm3+: YAG crystal in elevated temperatures and without compromising the memory bandwidth. Using various pumping and optimization techniques, we demonstrate multi-frequency window storage with a high memory bandwidth of 630 MHz. Moreover, we propose a general method for large-bandwidth atomic-frequency memory with non-Kramers rare-earth-ion (REI) in solids enabling significantly higher storage efficiency and bandwidth. Our study advances the practical applications of quantum memory devices based on REI-doped crystals. © 2024, CC BY.

关键词： Optical pumping

来源：评论

学校读者我要写书评

暂无评论

computer Vision-Based Hydroponic Lettuce Contour and Area Recognition System

Computer Vision-Based Hydroponic Lettuce Contour and Area Re...

引用

International Conference on electrical engineering and Informatics, ICEEI

作者： Samuel Raymond Pranowo Zener Lie Sukra Muhammad Zacky Asy'ari Marisa Paryasto Computer Engineering Department Automotive and Robotics Program BINUS ASO School of Engineering Bina Nusantara University Jakarta Indonesia Electrical Engineering Department Computer Engineering Program Telkom University Bandung Indonesia

Agriculture is a vital industry for the people of Indonesia, but there are several obstacles, including limited space in urban areas and inefficient conventional sorting and harvesting methods. Using computer vision to choose hydroponic lettuce that is ready to be harvested and a robotic arm to bring the selected lettuce yields, the produced system aims to solve this problem. The system's computer vision algorithms include color space conversion, intensity transformation, and an algorithm for following borders. This project's research methodology combines a quantitative approach with an experimental approach, which consists of conducting several trials on the built system. Several trials demonstrated that the computer vision software was able to accomplish the specified objectives. The average success rate of the developed computer vision system is 93%, while the average success rate of the robot arm is 85%.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Building a Brain computer Interface (BCI) Using Electroencephalogram (EEG) Signals' Classification

Building a Brain Computer Interface (BCI) Using Electroencep...

引用

International Conference on Advances in Biomedical engineering (ICABME)

作者： Mohammad Nabil Younis Sary Haj Sleiman Salma Khadra Amira Zaylaa Alaa Daher Mohammad Ayache Depatment of Electrical and Computer Engineering Beirut Arab University Biomedical Program Debbieh Lebanon

Development of Brain computer Interface (BCI) has been rapid since the mid 1990‘s. There are three criteria for BCI, (i) comfortability and possession of a suitable signal acquisition device, (ii) system validation and dissemination, and (iii) reliability and potentiality. As there are no BCI possessing the optimal criteria, it was essential to consider building a new one. Thereby, the paper investigates building BCI based on the utilization of EEG signals to translate brainwave patterns into actionable commands. The primary objective is to enhance communication capabilities for individuals afflicted with neurological disorders, empowering them to command external devices and engage more effectively with their surroundings. We built our model on EEG online dataset for the purpose of feature extraction and classification. Statistical features and Discrete Wavelet Transform (DWT) have been applied for feature selection. Multi-Layer Perceptron (MLP) and Radial Basis Function (RBF) were the classifiers involved. Results showed that the proposed architecture of MLP and RBF were able to classify the EEG signals into two classes (open eye and closed eye). Results also showed that the proposed approach, which is based on the combination of statistical features and DWT for features selection using AF3 and AF4 channels by the application of MLP, has 98% succession rate. BCI system based on Arduino circuit has been built after the classification Further algorithms and system evaluation need to be considered as future work.

关键词：

来源：评论

学校读者我要写书评

暂无评论

On the powerfulness of textual outlier exposure for visual OoD detection 23

On the powerfulness of textual outlier exposure for visual O...

引用

Proceedings of the 37th International Conference on Neural Information Processing Systems

作者： Sangha Park Jisoo Mok Dahuin Jung Saehyung Lee Sungroh Yoon Department of Electrical and Computer Engineering Seoul National University Department of Electrical and Computer Engineering Seoul National University and Interdisciplinary Program in Artificial Intelligence Seoul National University

Successful detection of Out-of-Distribution (OoD) data is becoming increasingly important to ensure safe deployment of neural networks. One of the main challenges in OoD detection is that neural networks output overconfident predictions on OoD data, make it difficult to determine OoD-ness of data solely based on their predictions. Outlier exposure addresses this issue by introducing an additional loss that encourages low-confidence predictions on OoD data during training. While outlier exposure has shown promising potential in improving OoD detection performance, all previous studies on outlier exposure have been limited to utilizing visual outliers. Drawing inspiration from the recent advancements in vision-language pre-training, this paper venture out to the uncharted territory of textual outlier exposure. First, we uncover the benefits of using textual outliers by replacing real or virtual outliers in the image-domain with textual equivalents. Then, we propose various ways of generating preferable textual outliers. Our extensive experiments demonstrate that generated textual outliers achieve competitive performance on large-scale OoD and hard OoD benchmarks. Furthermore, we conduct empirical analyses of textual outliers to provide primary criteria for designing advantageous textual outliers: near-distribution, descriptiveness, and inclusion of visual semantics. Code is available at https://***/wiarae/TOE

关键词：

来源：评论

学校读者我要写书评

暂无评论

On the Impact of Knowledge Distillation for Model Interpretability 40

On the Impact of Knowledge Distillation for Model Interpreta...

引用

40th International Conference on Machine Learning, ICML 2023

作者： Han, Hyeongrok Siwon, Kim Choi, Hyun-Soo Yoon, Sungroh Department of Electrical and Computer Engineering Seoul National University Seoul Korea Republic of Department of Computer Science and Engineering Seoul National University of Science and Technology Seoul Korea Republic of ZIOVISION Inc. Chuncheon Korea Republic of Interdisciplinary Program in Artificial Intelligence Seoul National University Seoul Korea Republic of

Several recent studies have elucidated why knowledge distillation (KD) improves model performance. However, few have researched the other advantages of KD in addition to its improving model performance. In this study, we have attempted to show that KD enhances the interpretability as well as the accuracy of models. We measured the number of concept detectors identified in network dissection for a quantitative comparison of model interpretability. We attributed the improvement in interpretability to the class-similarity information transferred from the teacher to student models. First, we confirmed the transfer of class-similarity information from the teacher to student model via logit distillation. Then, we analyzed how class-similarity information affects model interpretability in terms of its presence or absence and degree of similarity information. We conducted various quantitative and qualitative experiments and examined the results on different datasets, different KD methods, and according to different measures of interpretability. Our research showed that KD models by large models could be used more reliably in various fields. The code is available at https://***/Rok07/KD_***. © 2023 Proceedings of Machine Learning Research. All rights reserved.

关键词： Distillation

来源：评论

学校读者我要写书评

暂无评论

Fully Geometric Panoramic Localization

Fully Geometric Panoramic Localization

引用

Conference on computer Vision and Pattern Recognition (CVPR)

作者： Junho Kim Jiwon Jeong Young Min Kim Dept. of Electrical and Computer Engineering Seoul National University Dept. of Electrical Engineering Stanford University Interdisciplinary Program in Artificial Intelligence and INMC Seoul National University

ISBN: (数字)9798350353006

ISBN: (纸本)9798350353013

We introduce a lightweight and accurate localization method that only utilizes the geometry of 2D-3D lines. Given a pre-captured 3D map, our approach localizes a panorama image, taking advantage of the holistic 360° view. The system mitigates potential privacy breaches or domain discrepancies by avoiding trained or hand-crafted visual descriptors. However, as lines alone can be ambiguous, we express distinctive yet compact spatial contexts from relationships between lines, namely the dominant directions of parallel lines and the intersection between non-parallel lines. The resulting representations are efficient in processing time and memory compared to conventional visual descriptor-based methods. Given the groups of dominant line directions and their intersections, we accelerate the search process to test thousands of pose candidates in less than a millisecond without sacrificing accuracy. We empirically show that the proposed 2D-3D matching can localize panoramas for challenging scenes with similar structures, dramatic domain shifts or illumination changes. Our fully geometric approach does not involve extensive parameter tuning or neural network training, making it a practical algorithm that can be readily deployed in the real world. Project page including the code is available through this link: https://***/fgpl/.

关键词： Location awareness Geometry Training Visualization Accuracy Three-dimensional displays Pipelines

来源：评论

学校读者我要写书评

暂无评论

Automated Image Captioning with Multi-layer Gated Recurrent Unit 30

Automated Image Captioning with Multi-layer Gated Recurrent ...

引用

30th European Signal Processing Conference, EUSIPCO 2022

作者： Moral, Özge Taylan Kiliç, Volkan Onan, Aytug Wang, Wenwu Electrical and Electronics Engineering Graduate Program Izmir Katip Celebi University Turkey Department of Computer Engineering Izmir Katip Celebi University Turkey University of Surrey United Kingdom

ISBN: (纸本)9789082797091

Describing the semantic content of an image via natural language, known as image captioning, has recently attracted substantial interest in computer vision and language processing communities. Current image captioning approaches are mainly based on an encoder-decoder framework in which visual information is extracted by an image encoder and captions are generated by a text decoder, using convolution neural networks (CNN) and recurrent neural networks (RNN), respectively. Although this framework is promising for image captioning, it has limitations in utilizing the encoded visual information for generating grammatically and semantically correct captions in the RNN decoder. More specifically, the RNN decoder is ineffective in using the contextual information from the encoded data due to its limited ability in capturing long-term complex dependencies. Inspired by the advantage of gated recurrent unit (GRU), in this paper, we propose an extension of conventional RNN by introducing a multi-layer GRU that modulates the most relevant information inside the unit to enhance the semantic coherence of captions. Experimental results on the MSCOCO dataset show the superiority of our proposed approach over the state-of-the-art approaches in several performance metrics. © 2022 European Signal Processing Conference, EUSIPCO. All rights reserved.

关键词： Recurrent neural networks

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：