检索结果-内蒙古大学图书馆

arXiv 2024年

作者： Renduchintala, H.S.V.N.S. Kowndinya Bhatia, Sumit Ramakrishnan, Ganesh Media and Data Science Research Adobe Inc. India Dept. of Computer Science & Engineering Indian Institute of Technology Bombay India

Instruction Tuning involves finetuning a language model on a collection of instruction-formatted datasets in order to enhance the generalizability of the model to unseen tasks. Studies have shown the importance of balancing different task proportions during finetuning, but finding the right balance remains challenging. Unfortunately, there’s currently no systematic method beyond manual tuning or relying on practitioners’ intuition. In this paper, we introduce SMART (Submodular data Mixture strAtegy for instRuction Tuning) — a novel data mixture strategy which makes use of a submodular function to assign importance scores to tasks which are then used to determine the mixture weights. Given a fine-tuning budget, SMART redistributes the budget among tasks and selects non-redundant samples from each task. Experimental results demonstrate that SMART significantly outperforms traditional methods such as examples proportional mixing and equal mixing. Furthermore, SMART facilitates the creation of data mixtures based on a few representative subsets of tasks alone and through task pruning analysis, we reveal that in a limited budget setting, allocating budget among a subset of representative tasks yields superior performance compared to distributing the budget among all tasks. The code for reproducing our results is open-sourced at https://***/kowndinyarenduchintala/SMART. © 2024, CC BY.

关键词： Budget control

来源：评论

学校读者我要写书评

暂无评论

Gain Improvement for Single Band Ultra-Wideband Antenna by Slots Technique

Gain Improvement for Single Band Ultra-Wideband Antenna by S...

引用

IEEE SSD International Multi-Conference on Systems, Signals and Devices

作者： Ansam Q. Kamil Hasan M. Kadhim Riyam M. Alsammarraie Mobile Communications and Computing Engineering Dept. Engineering College University of Information Technology and Communications (UOITC) Baghdad Iraq Electrical Engineering Dept. College of Engineering Mustansiriyah University Baghdad Iraq Computer Engineering Techniques Dept. Madenat Alelem University College Baghdad Iraq

ISBN: (数字)9798350374131

ISBN: (纸本)9798350374148

The most important challenge in antenna design is increasing its gain because antenna gain is considered one of the main factors in determining antenna performance for applications in communication engineering. The paper tackles antenna enhancement by optimum antenna design for geometrical parameters of Rectangular-Shaped Slotted Single-Band Microstrip Patch Antenna. To increase the gain, the researchers used an optimization process of slot insertion in the antenna patch plane. The sizes of slots in the patch have been effectively changed to increase the gain. The design improved the directivity of the antenna as well. The dimensions of the patch for the array-shaped antenna are (21×12×1.6) mm. Its substrate dimensions are (30 x 30 x 1.6) mm. The design is tested by the frequencies 3.1 GHz-10.6 GHz Ultra-Wideband. For single-band 10.4 GHz, the proposed antenna gain is 5.06 dB, and VSWR is less than 2. The design is simulated successfully by computer simulation technology. The research technique causes gain enhancement with the range of 1.2 dB to 5.06 dB. The technique provides high gain at a single band for wireless communication applications.

关键词： Wireless communication Slot antennas Shape Patch antennas Microstrip antennas Ultra wideband antennas Gain

来源：评论

学校读者我要写书评

暂无评论

Knowledge Transfer Challenges: A Practical Study of Yemen Public Telecommunication Corporation

Knowledge Transfer Challenges: A Practical Study of Yemen Pu...

引用

Emerging Technologies for Dependable Internet of Things (ICETI), International Conference on

作者： Sadik Al-Taweel Ghassan Ahmed Ali Saleh Alhazbi Walid Yousef Belal Al-Fuhaidi Thamer Albater Essa Hezzam Abdullatif Ghallab Department of Information Systems University of Science and Technology Sana'a Yemen Faculty of Islamic Technology Universiti Islam Sultan Sharif Ali Brunei Darussalam Brunei Dept. of Computer Science & Eng Qatar University Doha Qatar Dept. of Information Technology University of Science and Technology Sana'a Yemen Department of Computer Science University of Science and Technology Sana'a Yemen Faculty of Business Open University Malaysia Sana'a Yemen Department of Information Systems Taibah University Medina Saudi Arabia

ISBN: (数字)9798331533557

ISBN: (纸本)9798331533564

The success of organizations today relies heavily on their intellectual capital, as knowledge is a key resource for value creation. Sharing and transferring knowledge enhances its value, embedding it within an organization’s processes and boosting competitiveness. This study focuses on the Yemen public telecommunication corporation (YPTC), which has adopted a knowledge-focused management approach. This study aims to explore challenges related to knowledge transfer in Yemen’s public sector, where such research is lacking. Using a descriptive and analytical method, data were collected via a questionnaire from 200 employees, yielding a response rate of 88%. The findings indicate a high level of knowledge transfer at YPTC, with individual, organizational, and technological factors identified as significant influences. The key barriers include trust, information technology support, motivation, and social networks.

关键词： Training Social networking (online) Organizations Boosting Communications technology Telecommunications Internet of Things information technology Knowledge transfer

来源：评论

学校读者我要写书评

暂无评论

Matthiola incana classification using CNN and class activation maps for model understanding

Matthiola incana classification using CNN and class activati...

引用

International Symposium on Computing and Networking Workshops (CANDARW)

作者： Kentaro Machida Soma Asuka Tomomi Osaka Fumikazu Watanabe Hironori Nakajo Dept. of Electrical Engineering and Computer Science Tokyo University of Agriculture and Technology Tokyo Japan Dept. of Horticultural Science Tokyo Metropolitan Mizuho Nogei High School Tokyo Japan

ISBN: (数字)9798331505349

ISBN: (纸本)9798331505356

In the production of seedlings for flowers and vegetables, selective differentiation is widely performed for only seedlings with high market-value prospects; however, many tasks require specialized knowledge. Mistakenly discarding seedlings with high market value leads to economic losses for farmers, making it crucial to provide a reliable system that also explains the basis for differentiation. This study proposes a method for the reliable classification of Matthiola incana (M. incana) by utilizing neural network-based image classification, supported by visualizing class activation maps to provide evidence for differentiation. Additionally, the study introduces the MiLIC (Matthiola incana Leaf Image Classification) dataset, developed for use in training the classification model. The dataset is available at https://***/datasets/Kentaro-Machida/Matthiola-incana-Leaf-Image-Classification.

关键词： Training Economics Visualization Accuracy Production Flowering plants Feature extraction Reliability Residual neural networks Image classification

来源：评论

学校读者我要写书评

暂无评论

Channel-Aware Domain-Adaptive Generative Adversarial Network for Robust Speech Recognition

arXiv

引用

arXiv 2024年

作者： Wang, Chien-Chun Chen, Li-Wei Chou, Cheng-Kang Lee, Hung-Shin Chen, Berlin Wang, Hsin-Min Dept. Computer Science and Information Engineering National Taiwan Normal University Taiwan Institute of Computer Science Academia Sinica Taiwan United-Link Co. Ltd. Taiwan

While pre-trained automatic speech recognition (ASR) systems demonstrate impressive performance on matched domains, their performance often degrades when confronted with channel mismatch stemming from unseen recording environments and conditions. To mitigate this issue, we propose a novel channel-aware data simulation method for robust ASR training. Our method harnesses the synergistic power of channel-extractive techniques and generative adversarial networks (GANs). We first train a channel encoder capable of extracting embeddings from arbitrary audio. On top of this, channel embeddings are extracted using a minimal amount of target-domain data and used to guide a GAN-based speech synthesizer. This synthesizer generates speech that faithfully preserves the phonetic content of the input while mimicking the channel characteristics of the target domain. We evaluate our method on the challenging Hakka Across Taiwan (HAT) and Taiwanese Across Taiwan (TAT) corpora, achieving relative character error rate (CER) reductions of 20.02% and 9.64%, respectively, compared to the baselines. These results highlight the efficacy of our channel-aware data simulation method for bridging the gap between source- and target-domain acoustics. © 2024, CC BY.

关键词： Generative adversarial networks

来源：评论

学校读者我要写书评

暂无评论

SEC-EnD: Stacked Ensemble Correlation-based Feature Selection Method for Emotion Detection

SEC-EnD: Stacked Ensemble Correlation-based Feature Selectio...

引用

2022 IEEE Silchar Subsection Conference, SILCON 2022

作者： Parui, Sricheta Samanta, Debasis Chakravorty, Nishant Indian Institute of Technology Advanced Technology Development Center Kharagpur721302 India Indian Institute of Technology Dept. of Computer Science and Engineering Kharagpur721302 India Indian Institute of Technology School of Medical Science and Technology Kharagpur721302 India

ISBN: (纸本)9781665471008

Because the brain signal is so complex, choosing the right features from the EEG signal has become a major area of research and is still important if you want to receive reliable findings. Due to the extremely high dimensionality of EEG data, redundant features and the dimensionality curse are frequently present, and the classification method frequently fails to produce the intended results. We have chosen an optimum feature set for emotion recognition using certain well-known feature selection techniques to address these issues. But most of them are unstable in nature, and for different training subsets, different types of feature selection methods work best. In this research, we provide SEC-END, a stacked ensemble correlation-based feature selection method for emotion detection that can combine and stack different feature selection methods to find the best features. This method has a 3-level feature selection technique, which helps to finetune the selection of features. After integrating the entire feature set, we took the features that each approach had in common and used the union quorum method to eliminate the features that were unnecessary. Then, using four classifiers-Random Forest, Decision Tree, SVM, and KNN-we tested the feature set. Regardless of the training set and classifier, the feature subset is outperforming other state-of-the-art techniques for all the dimensions. The experiment is also carried out with different sizes of windows for EEG signals so that the optimum window size can be recognized and used for further experiments. © 2022 IEEE.

关键词： Emotion Recognition

来源：评论

学校读者我要写书评

暂无评论

Efficient Load Frequency Regulation of Islanded Microgrids with Fractional-Order PI Controller 11

Efficient Load Frequency Regulation of Islanded Microgrids w...

引用

11th IEEE International Conference on Power Electronics, Drives and Energy Systems, PEDES 2024

作者： Bashishtha, T.K. Singh, V.P. Mathur, Akhilesh Murari, Krishna Mishra, Mrityunjay Kumar Al-Sumaiti, Ameena Saad M. National Institute of Technology RJ Jaipur302017 India The University of Toledo Dept. of Electrical Engineering and Computer Science OH43606 United States Khalifa University Smart Or Lab Apec Eecs Department Abu Dhabi United Arab Emirates

This article focuses on load frequency regulation of islanded microgrid (Iµ G), which consists of various distributed generation sources like photovoltaic cells, wind turbines, microturbine, biodiesel engine generators, and biogas turbine generators with energy storage systems such as the battery, flywheel, ultracapacitor, and electric vehicles. To eliminate undesirable frequency deviations and errors of (Iµ G), a fractional-order proportional-integral (FOPI) controller is utilized. Fine-tuning of FOPI controller gains is performed through a rule-based method. To prove the efficacy of FOPI controller in stabilizing frequency response, various simulation-based studies are conducted and compared. Moreover, comparative frequency response, with and without controller and transient specifications, underscore the importance of FOPI controller in stabilizing microgrid frequency. © 2024 IEEE.

关键词： Islanding

来源：评论

学校读者我要写书评

暂无评论

ReverBERT: A State Space Model for Efficient Text-Driven Speech Style Transfer

arXiv

引用

arXiv 2025年

作者： Brown, Michael Martinez, Sofia Singh, Priya Dept. of Computer Science University of Oregon United States Institute of AI Research Rochester Institute of Technology United States School of Information Technology IIT Kharagpur India

Text-driven speech style transfer aims to mold the intonation, pace, and timbre of a spoken utterance to match stylistic cues from text descriptions. While existing methods leverage large-scale neural architectures or pre-trained language models, the computational costs often remain high. In this paper, we present ReverBERT, an efficient framework for text-driven speech style transfer that draws inspiration from a state space model (SSM) paradigm, loosely motivated by the image-based method of Wang and Liu [14]. Unlike image domain techniques, our method operates in the speech space and integrates a discrete Fourier transform of latent speech features to enable smooth and continuous style modulation. We also propose a novel Transformer-based SSM layer for bridging textual style descriptors with acoustic attributes, dramatically reducing inference time while preserving high-quality speech characteristics. Extensive experiments on benchmark speech corpora demonstrate that ReverBERT significantly outperforms baselines in terms of naturalness, expressiveness, and computational efficiency. We release our model and code publicly to foster further research in text-driven speech style transfer. Copyright © 2025, The Authors. All rights reserved.

关键词： Discrete Fourier transforms

来源：评论

学校读者我要写书评

暂无评论

Single- and Multi-Hop BERT Question Classifier for Open-Domain Question Answering (SiMQC)

Single- and Multi-Hop BERT Question Classifier for Open-Doma...

引用

Iranian Conference on Electrical engineering (ICEE)

作者： Faeze Zakaryapour Sayyad Mahdi Bohlouli dept. Computer Science and Information Technology Institute for Advanced Studies in Basic Sciences (IASBS) Zanjan Iran

Multi-hop Question Answering has recently received particular attention in research and practice, especially in the context of conversational systems and answering complex questions. Various architectures have been proposed to answer these complex multi-hop questions. However, in real-world scenarios, a conversational system should answer both simple (single-hop) and complex (multi-hop) questions. In this work, we propose an efficient BERT question classifier that supports retrievers in the question-answering systems to process single- and multi-hop questions. We also released a mixed dataset consisting of both single- and multi-hop questions. We show that utilizing our classifier inside the QA system can improve these systems' accuracy and enable them to answer both kinds of questions considering their complexity.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Student's Performance Prediction Using Machine Learning Algorithms- A Comparative Study

Student's Performance Prediction Using Machine Learning Algo...

引用

2024 OPJU International technology Conference on Smart Computing for Innovation and Advancement in Industry 4.0, OTCON 2024

作者： Kadu, Varun Mishra, Pawan Kumar Dandhare, Sarthak Patni, Jagdish Chandra Chandel, Palash Pathak, Soham Bahadure, Nilesh Symbiosis Institute of Technology Siu Nagpur Campus Pune India Graphic Era Deemed to Be University Computer Science & Engineering Dehradun India Alliance University Alliance School of Advance Computing Dept. of Cse Karnataka Bengaluru India Nagpur Campus Pune India

ISBN: (纸本)9798350373783

Predicting and enhancing student performance has been a crucial topic of concentration in education amid the quick development of technology and the increasing need for higher-quality instruction. The use of machine learning technology to forecast pupils' academic success is covered in detail in this study report. This research project examined several prediction models that reliably predict students' performance according to a range of extracurricular and academic attributes. An overview of machine learning methods is given in this paper to automate the process of predicting students' academic performance and extracting outcomes. © 2024 IEEE.

关键词： Contrastive Learning

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：