检索结果-内蒙古大学图书馆

Proceedings of the 40th ACM/SIGAPP Symposium on Applied Computing

作者： Donggeun Ko Dongjun Lee Namjun Park Wonkyeong Shim Jaekwang Kim AiM Future Seoul Republic of Korea Maum AI Gyeonggi-do Republic of Korea Department of Computer Science and Engineering/Convergence Program for Social Innovation Sungkyunkwan University Suwon Republic of Korea Department of Computer Science and Engineering Sungkyunkwan University Suwon Republic of Korea School of Convergence / Convergence Program for Social Innovation Sungkyunkwan University Suwon Republic of Korea

ISBN: (纸本)9798400706295

Neural networks struggle with image classification when biases are learned and misleads correlations, affecting their generalization and performance. Previous methods require attribute labels (e.g. background, color) or utilizes Generative Adversarial Networks (GANs) to mitigate biases. We introduce DiffuBias, a novel pipeline for text-to-image generation that generates bias-conflict samples, without any training. By utilizing pretrained diffusion and image captioning models, DiffuBias generates, bias-conflict samples using the top-K losses from a biased classifier (fB) to debias the classifier. This method not only debiases effectively but also boosts classifier generalization capabilities. Our comprehensive experimental evaluations demonstrate that DiffuBias achieves state-of-the-art performance on benchmark datasets.

关键词： debiasing

来源：评论

学校读者我要写书评

暂无评论

FDAM: Filter-Dedicated Approximate Multiplier Design for Real-Time CNN Acceleration

引用

IEEE Transactions on computer-Aided Design of Integrated Circuits and Systems 2025年

作者： Nguyen, Kihwan Kim Xuan Truong Jeon, Dongsuk Lee, Hyuk-Jae Korea Republic of program Department of Electrical Engineering Seoul08826 Korea Republic of Inter-University Semiconductor Research Center Seoul National University Graduate School of Convergence Science and Technology Research Institute for Convergence Science Seoul08826 Korea Republic of Department of Electrical and Computer Engineering Seoul08826 Korea Republic of

Video super-resolution (VSR) is widely used in various high-definition applications, such as HDTVs and smartphones, requiring a dedicated upscaling technique for realtime full-HD generation. To reduce on-chip buffers for largesize output feature maps, a streaming VSR accelerator may employ an output stationary dataflow, leading to a large energy consumption caused by frequent filter switching. To mitigate this, we introduce a new filter-dedicated multiplier design for realtime VSR acceleration. We replace costly multipliers with adders, shifters, and multiplexers (MUXes), referred to as unified multiple constant multiplications (UMCM). The conventional UMCM, however, may incur considerable area/power overhead due to the unified topology constraint among different filter sets. To address this problem, we propose a new approximated MCM (AMCM) problem to relax the constraint and an approximate compatible graph synthesis (A-CGS) framework to efficiently solve AMCM by jointly searching for approximated filters and constructing a unified graph. Additionally, we suggest a lightweight finetuning method by freezing approximated filters and only finetuning biases, which can recover the original model’s accuracy within a few epochs. Experimental results with synthetic data demonstrate that AMCM reduces the area by up to 49.8%, 44.8%, and 40.3% when considering constant sets of 2, 4, and 8, respectively. Our designs with SR application achieve up to a 73.3% reduction in energy consumption. Experiments with the Set5 and Set14 datasets show that our model with bias correction achieves similar restoration performance compared to the eightbit models. © 1982-2012 IEEE.

关键词： Acceleration

来源：评论

学校读者我要写书评

暂无评论

LESA: Learnable LLM Layer Scaling-Up

arXiv

引用

arXiv 2025年

作者： Yang, Yifei Cao, Zouying Ma, Xinbei Yao, Yao Qin, Libo Chen, Zhi Zhao, Hai Department of Computer Science and Engineering Shanghai Jiao Tong University China Key Laboratory of Shanghai Education Commission for Intelligent Interaction and Cognitive Engineering Shanghai Jiao Tong University China Shanghai Key Laboratory of Trusted Data Circulation and Governance in Web3 China School of Computer Science and Engineering Central South University China ByteDance China

Training Large Language Models (LLMs) from scratch requires immense computational resources, making it prohibitively expensive. Model scaling-up offers a promising solution by leveraging the parameters of smaller models to create larger ones. However, existing depth scaling-up methods rely on empirical heuristic rules for layer duplication, which result in poorer initialization and slower convergence during continual pre-training. We propose LESA, a novel learnable method for depth scaling-up. By concatenating parameters from each layer and applying Singular Value Decomposition, we uncover latent patterns between layers, suggesting that inter-layer parameters can be learned. LESA uses a neural network to predict the parameters inserted between adjacent layers, enabling better initialization and faster training. Experiments show that LESA outperforms existing baselines, achieving superior performance with less than half the computational cost during continual pre-training. Extensive analyses demonstrate its effectiveness across different model sizes and tasks.1 Copyright © 2025, The Authors. All rights reserved.

关键词： Multilayer neural networks

来源：评论

学校读者我要写书评

暂无评论

Convergence of energy-based learning in linear resistive networks

arXiv

引用

arXiv 2025年

作者： Huijzer, Anne-Men Chaffey, Thomas Besselink, Bart van Waarde, Henk J. Bernoulli Institute for Mathematics Computer Science and Artificial Intelligence University of Groningen Groningen Netherlands CogniGron - Groningen Cognitive Systems and Materials Center Netherlands Control Group Department of Engineering University of Cambridge United Kingdom School of Electrical and Computer Engineering University of Sydney Australia

Energy-based learning algorithms are alternatives to backpropagation and are well-suited to distributed implementations in analog electronic devices. However, a rigorous theory of convergence is lacking. We make a first step in this direction by analysing a particular energy-based learning algorithm, Contrastive Learning, applied to a network of linear adjustable resistors. It is shown that, in this setup, Contrastive Learning is equivalent to projected gradient descent on a convex function, for any step size, giving a guarantee of convergence for the *** Codes 65K10, 68T05, 93B30, 93D99 © 2025, CC BY-SA.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

ATP_mCNN: Predicting ATP binding sites through pretrained language models and multi-window neural networks

引用

computers in Biology and Medicine 2025年 185卷 109541-109541页

作者： Le, Van-The Malik, Muhammad-Shahid Lin, Yi-Jing Liu, Yu-Chen Chang, Yan-Yun Ou, Yu-Yen Department of Computer Science and Engineering Yuan Ze University Chung-Li32003 Taiwan Graduate Program in Biomedical Informatics Yuan Ze University Chung-Li32003 Taiwan Department of Computer Sciences Karakoram International University Gilgit-Baltistan 15100 Pakistan

Adenosine triphosphate plays a vital role in providing energy and enabling key cellular processes through interactions with binding proteins. The increasing amount of protein sequence data necessitates computational methods for identifying binding sites. However, experimental identification of adenosine triphosphate-binding residues remains challenging. To address the challenge, we developed a multi-window convolutional neural network architecture taking pre-trained protein language model embeddings as input features. In particular, multiple parallel convolutional layers scan for motifs localized to different window sizes. Max pooling extracts salient features concatenated across windows into a final multi-scale representation for residue-level classification. On benchmark datasets, our model achieves an area under the ROC curve of 0.95, significantly improving on prior sequence-based models and outperforming convolutional neural network baselines. This demonstrates the utility of pre-trained language models and multi-window convolutional neural networks for advanced sequence-based prediction of adenosine triphosphate-binding residues. Our approach provides a promising new direction for elucidating binding mechanisms and interactions from primary structure. © 2024 Elsevier Ltd

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Rule-Based Dialect of Tulang Bawang Stemmer

Rule-Based Dialect of Tulang Bawang Stemmer

引用

Advancement in Data science, E-learning and Information Systems (ICADEIS), International Conference on

作者： Zaenal Abidin Akmal Junaidi Wamiliana Favorisen Rosyking Lumbanraja Dian Kurniasari Rohmat Indra Borman Doctoral Program of Mathematics and Natural Sciences Universitas Lampung Bandar Lampung Indonesia Department of Computer Science Universitas Lampung Bandar Lampung Indonesia Department of Mathematics Universitas Lampung Bandar Lampung Indonesia Faculty of Engineering and Computer Science Universitas Teknokrat Indonesia Bandar Lampung Indonesia

ISBN: (数字)9798331513320

ISBN: (纸本)9798331513337

Stemming, an essential procedure in natural language processing (NLP), diminishes words to their base forms, facilitating tasks such as information retrieval and sentiment analysis. Although stemming techniques for highresource languages are well-developed, numerous low-resource languages, including dialect of Tulang Bawang, suffer from inadequate solutions owing to a scarcity of linguistic data and resources. Existing systems, including rule-based stemmers, have demonstrated efficacy in processing low-resource languages such as Indonesian and Javanese by utilizing established morphological rules. Nonetheless, these methods encounter considerable obstacles, such as restricted adaptability, inability to accommodate unusual root structures, and excessive dependence on fixed rules that might result in over- or understemming. Rule-based methodologies frequently misidentify roots when faced with intricate affixes or unconventional word forms. We introduce an improved rule-based Tulang Bawang Stemmer aimed at overcoming these constraints by enhancing current linguistic rules and integrating new patterns specific to the language's morphology. Assessed on 500 test samples and 200 independent test samples, our improved stemmer attained gold standard evaluation metrics of 96.2% and 94%, respectively, surpassing prior implementations in both precision and generalization. The findings demonstrate the potential of enhanced rule-based techniques to improving NLP for lowresource languages. Improved stemming performance enables better downstream applications, promotes more efficient text analysis, and advances research in underrepresented languages.

关键词： Measurement Sentiment analysis Text analysis Electronic learning Morphology Linguistics Data science Information retrieval Machine translation Information systems

来源：评论

学校读者我要写书评

暂无评论

Ontologies in Design: How Imagining a Tree Reveals Possibilities and Assumptions in Large Language Models 25

Ontologies in Design: How Imagining a Tree Reveals Possibili...

引用

Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems

作者： Nava Haghighi Sunny Yu James A. Landay Daniela Rosner Computer Science Stanford University Stanford California USA Symbolic Systems Program Stanford University Stanford California USA Computer Science Department Stanford University Stanford California USA Human Centered Design & Engineering University of Washington Seattle Washington USA

来源：评论

学校读者我要写书评

暂无评论

Auto Digit Selection for Most Significant Digit First Multiplication

引用

IEEE Embedded Systems Letters 2025年

作者： Gorgin, Saeid Jaberipur, Ghassem Lee, Jeong-A Shin, Seokjoo Kim, Jungrae Computer Engineering Iranian Research Organization for Science and Technology Tehran Iran Semiconductor Systems Engineering Department College of Information and Communication Engineering SungKyunKwan University Korea Republic of Brain Pool Program for Chosun University Gwangju Korea Republic of Shahid Beheshti University Tehran Iran Chosun University Department of Computer Engineering Gwangju Korea Republic of

Performance in modern embedded systems, particularly those executing computation-intensive signal/image processing and machine learning algorithms, is critically dependent on the efficiency of multiplication operations. Serial multiplication offers key advantages in these resource-constrained environments, including low memory usage and minimal interconnect complexity. The most significant digit first (MSDF) serial multiplication scheme further enhances efficiency by enabling early termination and supporting variable precision. It is ideal for embedded applications where power and performance are critical. This paper proposes an auto-selection scheme for MSDF multiplication, where product digits are generated as a byproduct of residual maintenance, leading to speed, cost, and power efficiency improvements. We demonstrate the implementation of serial-serial and serial-parallel multipliers, where the multiplier operand in the latter is fully available throughout the computation. Simulation and synthesis results, compared to the most relevant previous works, show reductions in delay (16% and 20%), area consumption (11% and 23%), and power dissipation (7% and 20%), making our approach well-suited for embedded systems requiring efficient, low-power multiplication hardware. © 2009-2012 IEEE.

关键词： Left-to-Right Arithmetic Machine Learning Algorithms MSDF Multiplication Online Arithmetic

来源：评论

学校读者我要写书评

暂无评论

Development of a Phase-Locked Loop Circuit in Running Frequency Mode for Capacitance-to-Voltage Conversion

Development of a Phase-Locked Loop Circuit in Running Freque...

引用

International Electrical engineering Congress (iEECON)

作者： Tada On-khong Noppadon Sisuk Paramote Wardkein Ratchanoo Katman Kriangsak Prompak Department of Physics Faculty of Science Naresuan University Phitsanulok Thailand Program in Computer Engineering Faculty of Industrial Technology Phibunsongkhram Rajabhat University Phitsanulok Thailand Department of Telecommunication School of Engineering King Mongkut's Institute of Technology Ladkrabang Bangkok Thailand Program of Physics Faculty of Science and Technology Phibunsongkhram Rajabhat University Phitsanulok Thailand

ISBN: (数字)9798331543952

ISBN: (纸本)9798331543969

Capacitance-to-voltage conversion is essential for capacitive sensors in various industries, including touch interfaces, medical devices, and moisture measurement. However, circuit design faces challenges like non-linearity, noise, and small capacitance variations affecting voltage signals. This article proposes a phase-locked loop in free-running oscillator mode with a frequency-to-voltage converter to enhance accuracy and stability. Test results show a 97.9% measurement accuracy, demonstrating reduced noise and improved stability. The proposed circuit is ideal for high-precision applications in diverse environments, overcoming limitations of traditional methods.

关键词： Accuracy Sensitivity Moisture measurement Noise Moisture Capacitance Frequency conversion Circuit stability Phase locked loops Standards

来源：评论

学校读者我要写书评

暂无评论

System for Automating Medication Orders to the Hospital Pharmacy by Nurses in Peru Through Voice Recognition

System for Automating Medication Orders to the Hospital Phar...

引用

Software engineering and computer science (CSECS), International Conference on

作者： Conrado Moreno-Fernández Fabian Colque-Fernández José Luis Castillo-Sequera Lenis Wong Information Systems Engineering Program Universidad Peruana de Ciencias Aplicadas Lima Perú Department of Computer Science Universidad de Alcalá Alcalá de Henares Spain

ISBN: (数字)9798331522216

ISBN: (纸本)9798331522223

The management of nurses' orders to a Peruvian hospital pharmacy faces efficiency challenges, which affects time spent on patient care. This context highlights the need for a system that optimizes this process. In this sense, a web-based system is proposed for the automation of orders from the nurse's cell phone, which uses a customized model for automatic speech recognition. The development of the system was divided into three stages: first, data collection with 21 orders placed by three nurses and 978 audio recordings to validate the automatic speech recognition model; second, implementation of the system allowing agile order processing; and finally, evaluation of results in terms of time and accuracy. The findings indicate that the processing time ranged from 37.70 to 60.15 seconds, representing a 97%-time reduction in operational tasks. In addition, the Word Error Rate was 10.51%, significantly lower than the 25.05% of Whisper Large V2. These results demonstrate the system's potential for adoption in various hospital contexts.

关键词： Automation Accuracy Hospitals Error analysis Terminology Transforms Time measurement Artificial intelligence Automatic speech recognition Software engineering

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：