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

A smart study on medicinal plants identification and classification using image processing techniques 3

A smart study on medicinal plants identification and classif...

引用

3rd International Conference on Intelligent Communication Technologies and Virtual Mobile Networks, ICICV 2021

作者： Pukhrambam, Banita Rathna, R. Vels Institute of Science Technology Advanced Studies Dept. of Computer Science and Engineering Chennai India

ISBN: (纸本)9780738111834

Plants play an important role in human life for providing oxygen, food, housing, medicine, energy, housing, and environmental protection. Plants are rich in medicinal esteem and contain dynamic elements for medicinal use because of the global warming populace, lack of expert help for research, lack of government upholding research exercises, and familiarity with medicinal plants. Numerous utility plants are getting pulverized. Manual identification of plants requires significant investment and requires the identification of plants with the help of experts. To address this issue, individuals need to acquire a more prominent advantage in robotized identification and medicinal plant classification. The research region's image process field is dynamic in the automatic identification and classification of medicinal plants. Feature extraction and classification are the important developments in identifying medicinal plants that affect the overall accuracy of the system for scientific categorization. This research examines recent methods of image processing for the classification of medicinal plants. © 2021 IEEE.

关键词： Housing

来源：评论

学校读者我要写书评

暂无评论

Pre-Trained Language Models for Keyphrase Prediction: A Review

arXiv

引用

arXiv 2024年

作者： Umair, Muhammad Sultana, Tangina Lee, Young-Koo Dept. of Computer Science and Engineering Kyung Hee University Global Campus. Yongin-si Korea Republic of Dept. of Electronics and Communication Engineering Hajee Mohammad Danesh Science and Technology University Bangladesh

Keyphrase Prediction (KP) is essential for identifying keyphrases in a document that can summarize its content. However, recent Natural Language Processing (NLP) advances have developed more efficient KP models using deep learning techniques. The limitation of a comprehensive exploration jointly both keyphrase extraction and generation using pre-trained language models spotlights a critical gap in the literature, compelling our survey paper to bridge this deficiency and offer a unified and in-dept. analysis to address limitations in previous surveys. This paper extensively examines the topic of pre-trained language models for keyphrase prediction (PLM-KP), which are trained on large text corpora via different learning (supervisor, unsupervised, semi-supervised, and self-supervised) techniques, to provide respective insights into these two types of tasks in NLP, precisely, Keyphrase Extraction (KPE) and Keyphrase Generation (KPG). We introduce appropriate taxonomies for PLM-KPE and KPG to highlight these two main tasks of NLP. Moreover, we point out some promising future directions for predicting keyphrases. © 2024, CC BY.

关键词：

来源：评论

学校读者我要写书评

暂无评论

An Analysis and Investigation of InfoStealers Attacks during COVID'19: A Case Study 2

An Analysis and Investigation of InfoStealers Attacks during...

引用

2nd International Conference on Secure Cyber Computing and Communications, ICSCCC 2021

作者： Sharma, Ravi Sharma, Nonita Mangla, Monika Dr. B. R. Ambedkar National Institute of Technology Dept. of Computer Science and Engineering Jalandhar India Lokmanya Tilak College of Engineering Dept. of Computer Science and Engineering Navi Mumbai India

ISBN: (纸本)9781665444156

A COVID-19 pandemic is a rare event that distorted the lives of many civilians worldwide stemming from what is commonly referred to as the new-normal regarding communal standards and the way we live and work. Apart from the devastating influence on humanity and business, the pandemic also spawned a set of unprecedented cyber-attacks. These cyber-attacks during such critical time have traumatized the society. The rise in cyber-attacks is triggered due to rising apprehensions due to this pandemic like never before. Resultantly, a significant increase in the quantity and variety of cyber-attacks is observed since emergence of Covid-19. Cybercriminals promptly leveraged this pandemic premise to rebrand general attack vectors. These attacks are typically info stealers and vary from attacks of minimal intricacy such as 404 Keylogger to the latest and more frequent attacks such as Lokibot. This manuscript analyses the COVID-19 pandemic from the perspective of cyber-attack and establishes that a huge surge has been recorded in the cyberattacks during this pandemic across the globe. In this manuscript, authors analyze the cyberattacks from the perspective of critical macro events to divulge the modus-operandi of cybercrimes. This analysis helps to employ the USA as a case study to determine how cyber-criminals gained access to vital information related to significant events. As a result, this also helps to create and project cyber-crime campaigns meticulously. © 2021 IEEE.

关键词： COVID-19

来源：评论

学校读者我要写书评

暂无评论

Whisper: A Curious Case of Valid and Employed Mallory in Cloud Computing 8

Whisper: A Curious Case of Valid and Employed Mallory in Clo...

引用

8th IEEE International Conference on Cyber Security and Cloud Computing and 2021 7th IEEE International Conference on Edge Computing and Scalable Cloud, CSCloud-EdgeCom 2021

作者： Patgiri, Ripon National Institute of Technology Silchar Dept. Computer Science and Engineering Assam Cachar788010 India

ISBN: (纸本)9781665443777

Apart from Security and privacy, there are crucial requirements of secrecy for modern practice in Cloud Computing. Secrecy protocol keeps client's data secret, and it protects the secrets from being misused by any other unintended users. It is the essential requirement to establish trust among users, vendors, and countries. However, the most modern system does not provide secrecy of user and user's data. Therefore, the secrecy of the user and the user's data remains no more secret. It is exposed intentionally or unintentionally. Moreover, there is no data secrecy for administrators in Cloud Computing such that the administrators cannot easily misuse the data. Therefore, trust in the company needs to be restored by devising a new protocol. Hence, we propose a secrecy protocol for establishing trust among the users, vendors, and countries, called Whisper, which maintains hard secrecy protocol. Also, this secrecy protocol follows "only by my choice"philosophy. This paper demonstrates how to achieve this protocol to re-establish trust among users, vendors, and countries in Cloud Computing. Whisper uses asymmetric cryptography protocol, which requires random numbers and prime numbers. We have tested the randomness of generated pseudo-random numbers of Whisper in NIST SP 800-22 and passes all 15 statistical tests of randomness for various input. This paper also presents the generation of a randomly chosen prime number and a random number, and demonstrates how to reproduce the randomly chosen prime number and the generated random number to eliminate administrators from list of intended user's list. © 2021 IEEE.

关键词： Cloud computing

来源：评论

学校读者我要写书评

暂无评论

Deep Learning-Based Facial Emotion Recognition for Advanced Human-computer Interaction 1

Deep Learning-Based Facial Emotion Recognition for Advanced ...

引用

1st International Conference on Advances in computer Science, Electrical, Electronics, and Communication Technologies, CE2CT 2025

作者： Chindiyababy, U. Kakkar, Preetish Vedula, Jagadeesh Yunus, Jumaniyozov Umidbek, Abdalov Sharma, Subhash SRM Valliammai Engineering College Dept. of Information Technology Chennai India India Digital Transformation Lead Meghaz Inc. HoustonTX United States Urgench state university Department of psychology and pedagogy Urgench Uzbekistan Mamun university Department of History Khiva Uzbekistan Tula's Institute Department of Computer Science and Engineering Dehradun India

ISBN: (纸本)9798331518578

Deep learning technology establishes advanced methods for enhancing Human-computer Interaction systems by providing machines with the ability to sense and understand human emotional states. The capability's value stems from its applications, including healthcare situations alongside virtual assistants, gaming interfaces, and educational platforms seeking personalized functions. We present a deep learning framework that unites multimodal data fusion with Convolutional Neural Network(CNN) to enhance emotion recognition systems while improving performance and interpretability. The proposed framework undergoes testing against three benchmark emotion datasets: FER2013, AffectNet, and CK+ dataset. Our experimental findings show that the proposed system delivers a 92.5% accuracy rate for emotion recognition which exceeds traditional unimodal models by a maximum of 12% improvement. Context-aware modeling achieves a 15% performance boost during real-world applications while lightweight architectural design reduces system processing requirements by 25% to support edge device runtime operations. Through explainable AI integration users can see how the model functions to make decisions therefore establishing both trust and ethical practices in use. This research delivers substantial progress in emotion recognition study through an approach that establishes a deployable system that maintains both operational effectiveness and ethical operations across multiple Human-computer Interaction application domains. © 2025 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Virtual Hands: Real Time Keyboard, Desktop & Application Navigation using Gestures

Virtual Hands: Real Time Keyboard, Desktop & Application Nav...

引用

Fourth Industrial Revolution Based technology and Practices (ICFIRTP), International Conference on

作者： Harshit Sharma Richa Saxena Satendra Kumar Ashish Saini Mohammad Saad Mohammad Faraz Dept. of Computer Science and Engineering Quantum University Roorkee India Dept. of Computer Science and Engineering Moradabad Institute of Technology Moradabad India

In computer system keyboard is the most prominent input medium of all time. But lately, human community is living in an era of global pandemic being afraid of suffering from Coronavirus (Covid-19) and hence each and every person avoids touching anything. This is because of the fear of contracting this contagious virus and their mutants. So, to mitigate this issue, we present a method "webcam based virtual keyboard interface" to interact with a computer system. The code of this method is written using pre- built modules like OpenCV, MediaPipe, PyVDA, Win32API, etc. and Python 3.9. This approach uses matching the index finger and middle finger on the specific key. After that the virtual desktop switching mechanism is done by PyVDA. The PyttSX3 library plays the sound whenever any key is pressed or when a desktop switch is initiated, corresponding to the key pressed or the desktop switched. In this approach no additional hardware device other than the webcam, that is already available in the system, is required. This approach is also useful for those persons who wants the access the system, even when their hands are dirty.

关键词： COVID-19 Pathogens Navigation Webcams Pandemics Fingers Keyboards

来源：评论

学校读者我要写书评

暂无评论

Blockchain technology to Secure Bluetooth

arXiv

引用

arXiv 2022年

作者： Kalogiratos, Athanasios Kantzavelou, Ioanna Dept. of Informatics and Computer Engineering University of West Attica Egaleo Athens12243 Greece

Bluetooth is a communication technology used to wirelessly exchange data between devices. In the last few years there have been found a great number of security vulnerabilities, and adversaries are taking advantage of them causing harm and significant loss. Numerous system security updates have been approved and installed in order to sort out security holes and bugs, and prevent attacks that could expose personal or other valuable information. But those updates are not sufficient and appropriate and new bugs keep showing up. In Bluetooth technology, pairing is identified as the step where most bugs are found and most attacks target this particular process part of Bluetooth. A new technology that has been proved bulletproof when it comes to security and the exchange of sensitive information is Blockchain. Blockchain technology is promising to be incorporated well in a network of smart devices, and secure an Internet of Things (IoT), where Bluetooth technology is being extensively used. This work presents a vulnerability discovered in Bluetooth pairing process, and proposes a Blockchain solution approach to secure pairing and mitigate this vulnerability. The paper first introduces the Bluetooth technology and delves into how Blockchain technology can be a solution to certain security problems. Then a solution approach shows how Blockchain can be integrated and implemented to ensure the required level of security. Certain attack incidents on Bluetooth vulnerable points are examined and discussion and conclusions give the extension of the security related problems. © 2022, CC BY-NC-SA.

关键词： Bluetooth

来源：评论

学校读者我要写书评

暂无评论

Improve the Efficiency and Security by Digitalizing the Sri Lankan Police Department 3

Improve the Efficiency and Security by Digitalizing the Sri ...

引用

3rd International Conference on Advancements in Computing, ICAC 2021

作者： Yapa, K. Senaratne, A.N. Pathirathna, R.P.D.T.D. Priyamantha, Y.P.C.N. Dias, W.D.S. Katukithulgala, B. Sri Lanka Institute of Information Technology Dept. of Computer System Engineering Malabe Sri Lanka

ISBN: (纸本)9781665408622

Police, as the main law enforcement organization in Sri Lanka, play a major role in preserving the public order, preventing criminal, terrorist activities, and enforcing the law. As a developing nation, Sri Lanka Police Department is moving forward with emerging technologies to effectively manage operations and rapidly respond to opportunities and challenges. Compared to developed countries, Police Department is still far more behind when working with modern technologies. Lack of coordination between departments, duplication of data, delay of services, unauthorized data modification, absence of data analysis and prediction systems, lack of transparency, and corruption were identified as the main issues that exist in the police department. In this paper, we propose 4 main improvements which help to overcome those issues and challenges. Improving the existing police clearance certificate system to issue digital certificates capable of verifying the authenticity, integrity of the document and providing efficient delivery and great user experience to the applicants. Implementing a Blockchain network to store confidential data securely by providing transparency and integrity. Developing a machine learning model to predict crimes geographically and predict fight crime action based on data gathered from a real-Time camera system (CCTV). © 2021 IEEE.

关键词： Forecasting

来源：评论

学校读者我要写书评

暂无评论

A Non-destructive Short Circuit Withstand Time Screening Methodology for Commercially Available SiC Power MOSFET

A Non-destructive Short Circuit Withstand Time Screening Met...

引用

IEEE Workshop on Wide Bandgap Power Devices and Applications (WiPDA)

作者： Monikuntala Bhattacharya Michael Jin Jiashu Qian Limeng Shi Shiva Houshmand Hengyu Yu Marvin H. White Anant K. Agarwal Atsushi Shimbori Aditya Aman Nikhil Bhardwaj Dept. of Electrical & Computer Engineering The Ohio State University Columbus Ohio USA Ford Motor Company Deaborn MI USA Dept. of Electrical Engineering The Indian Institute of Technology Bombay Mumbai India

ISBN: (数字)9798350375602

ISBN: (纸本)9798350375619

This work presents a novel non-destructive methodology to screen out devices with lower short circuit withstand time (SCWT) in commercial $4 \mathrm{H}-\mathrm{SiC}$ power metal-oxide semiconductor field effect transistors (MOSFETs) featuring a planar gate structure. The proposed screening process can effectively identify and remove devices with lower SCWT from a batch of devices with the same lot number without degrading the performance of the reliable ones.

关键词： Performance evaluation MOSFET Silicon carbide Measurement techniques MOSFET circuits Logic gates Reliability engineering Object recognition Integrated circuit reliability Protection

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

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

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：