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

I. Predicting future citation counts using machine learning 22

I. Predicting future citation counts using machine learning

引用

22nd International Arab Conference on Information Technology, ACIT 2021

作者： Mansour, Khalid Al-Daoud, Essam Al-Karaky, Baha Data Science and AI Dept. Zarqa University Zarqa Jordan Computer Science Dept. Zarqa University Zarqa Jordan

ISBN: (纸本)9781665419956

The number of citations for a research paper is an important criterion in the scientific research community and is also an important quality indicator for both the author and the affiliated institution. Since the number of researchers, universities and scientific institutes are increasing, the number of submitted papers to scientific journals increased substantially. Consequently, the time needed to process the submitted papers is increased as well. This results in delaying publishing high quality papers. This paper proposes an approach to predict the future citations of submitted research papers using machine learning algorithms to speed up the processing time of the predicted high quality submitted papers. A dataset solicited from the International Arab Journal of Information Technology (IAJIT) is used in our experiments. The dataset covers the last ten years. Sixteen machine learning algorithms are trained on the dataset. Two sets of results are produced. The first one is related to the relative importance of features used in the prediction process. The second set shows the results of the sixteen machine learning algorithms used to predict future citations of submitted papers. The experimental results show that the number of references is the most important feature while the number of authors shown on papers is the least important feature. In addition, the results produced by neural network and voting classifier 1 techniques are slightly better than other techniques in predicting future citations. Naïve base comes next. The rest of the used machine learning methods show similar performance. © 2021 IEEE.

关键词： Forecasting

来源：评论

学校读者我要写书评

暂无评论

Analysis and Ranking Resume using Machine Learning Algorithms and NLP

Analysis and Ranking Resume using Machine Learning Algorithm...

引用

Advances in Computing, Communication and Networking (ICAC2N), International Conference on

作者： Pratik Chandrasen Deshmukh M. R. Bendre Dept. of Data Science Engineering Pravara Rural Engineering College Loni India Dept. of Computer Engineering Pravara Rural Engineering College Loni India

ISBN: (数字)9798350356816

ISBN: (纸本)9798350356823

Specialized recruiting firms have emerged as a result of the expansion of the Indian employment market. These agencies use machine learning models to expedite the hiring process and provide their clients with the best candidates. Thus, the suggested methodology offers the potential to improve the hiring process by incorporating the advantages of machine learning in addition to natural language processing(NLP). Employing NLP techniques, one may accurately grade resumes according to the company's criteria and extract useful information from them. After thorough testing and improvement, the web portal might prove to be an invaluable tool for hiring managers and job seekers ***, one might think about adding features to the CV classifier and ranker, such as customized feedback and recommendations for enhancements that might assist candidates in understanding where they might be lacking and increase their chances of getting recruited.

关键词： Machine learning algorithms Resumes Employment Natural languages Machine learning Software Portals Recruitment Testing Qualifications

来源：评论

学校读者我要写书评

暂无评论

Remote Sensing Aircraft Classification Harnessing Deep Learning Advancements

Remote Sensing Aircraft Classification Harnessing Deep Learn...

引用

International Conference on Emerging Technologies, ICET

作者： Ahmad Saeed Haasha Bin Atif Usman Habib Mohsin Bilal Dept. of AI & Data Science National University of Computer and Emerging Sciences Islamabad Pakistan

Remote sensing imagery is challenging to analyze due to its diverse sources, object image variability and contextual backgrounds. In current era, Aviation Industry is continuously progressing from specific domain of aircraft classification to improving in its safety measures, maintenance procedures and diverse flight operations by utilizing the potential of satellite imagery and Deep Learning methods. This paper highlights and explores the potential of Transfer Learning with state-of-the-art Deep Learning architectures on publicly available Multi Type Aircraft Remote Sensing Imagery datasets. In the same context, our experiments and training mechanism on the state-of-the-art models, like Vision Transformers, ResNet50v2, EfficientNetB0 and InceptionNetV3 outperforms earlier methods on the benchmark remote sensing military aircraft dataset. Our experiments on mentioned Deep learning models yielded, 95%, 93.4%, 92% and 80% classification accuracy due to advance architecture design and fine tuning techniques by allowing them to capture more intricate features and patterns of multifaceted military platforms. This research will provide a valuable resource in the aviation industry in the perspective of aircraft recognition domain by utilizing the power of advance Deep Learning methods to handle the complexity and diversity of remote sensing imagery.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Windmill Pitch Prediction And Error Correction With Adaptive Control And Current Generation Analysis Using data Analysis

Windmill Pitch Prediction And Error Correction With Adaptive...

引用

Energy, Materials and Communication Engineering (ICEMCE), International Conference on

作者： M. S. Minu Rohit Suryaa S Jovan Titus J Sharan S Dept. of Computer Science Engineering SRM Institute of Science and Technology Chennai India Dept. Computer Science Engineering Specialization With Big Data Analytics SRM Institute of Science and Technology Chennai India

The growing prominence of wind energy as an eco-friendly power alternative amplifies the need for optimal turbine performance. Central to this is the ability to accurately predict and modulate the wind turbine pitch. Our research focuses on integrating adaptive control strategies with advanced power generation analytics to enhance pitch prediction precision. We have processed and analyzed vast amounts of wind data using state-of-the-art data analytics to discern patterns in wind velocity changes accurately. The continuous stream of real-time data is the backbone of our adaptive controls, ensuring perpetual pitch adjustments that optimize energy capture. In tandem with this, we present a thorough analysis linking momentary wind shifts to power production, ensuring consistent and efficient power output. The culmination of these methodologies offers a breakthrough solution in wind energy optimization. Preliminary results highlight the paramount importance of real-time data analytics in revolutionizing green energy advancements.

关键词：

来源：评论

学校读者我要写书评

暂无评论

ChatGPT-guided Semantics for Zero-shot Learning

ChatGPT-guided Semantics for Zero-shot Learning

引用

2023 International Conference on Digital Image Computing: Techniques and Applications, DICTA 2023

作者： Shubho, Fahimul Hoque Chowdhury, Townim Faisal Cheraghian, Ali Saberi, Morteza Mohammed, Nabeel Rahman, Shafin North South University Dept. of Electrical and Computer Engineering Bangladesh University of Adelaide Australian Institute for Machine Learning Australia Data61 Commonwealth Scientific and Industrial Research Organisation Australia University of Technology Sydney School of Computer Science and Dsi Australia

ISBN: (纸本)9798350382204

Zero-shot learning (ZSL) aims to classify objects that are not observed or seen during training. It relies on class semantic description to transfer knowledge from the seen classes to the unseen classes. Existing methods of obtaining class semantics include manual attributes or automatic word vectors from language models (like word2vec). We know attribute annotation is costly, whereas automatic word vectors are relatively noisy. To address this problem, we explore how ChatGPT, a large language model, can enhance class semantics for ZSL tasks. ChatGPT can be a helpful source to obtain text descriptions for each class containing related attributes and semantics. We use the word2vec model to get a word vector using the texts from ChatGPT. Then, we enrich word vectors by combining the word embeddings from class names and descriptions generated by ChatGPT. More specifically, we leverage ChatGPT to provide extra supervision for the class description, eventually benefiting ZSL models. We evaluate our approach on various 2D image (CUB and AwA) and 3D point cloud (ModelNet10, ModelNet40, and ScanObjectNN) datasets and show that it improves ZSL performance. Our work contributes to the ZSL literature by applying ChatGPT for class semantics enhancement and proposing a novel word vector fusion method. © 2023 IEEE.

关键词： Zero-shot learning

来源：评论

学校读者我要写书评

暂无评论

Youth Unemployment Mitigation: Leveraging Machine Learning and Exploratory data Analysis for Evidence-Driven Strategies

Youth Unemployment Mitigation: Leveraging Machine Learning a...

引用

Engineering Informatics, IEEE

作者： S. Aarthi Dharmita A Pragathi G S Elakkiya Dept. of Computer Science Engineering SRM Institute of Science and Technology Chennai India Dept. Computer Science Engineering specialization with Big Data Analytics SRM Institute of Science and Technology Chennai India

ISBN: (数字)9798350338522

ISBN: (纸本)9798350338539

Youth unemployment represents an urgent global socioeconomic challenge. This study combines machine learning and exploratory data analysis to understand and solve this problem comprehensively. Traditional approaches often rely on qualitative assessments and historical data requiring a data-driven transition to more effective solutions. Our approach integrates machine learning algorithms and Exploratory data Analysis techniques to analyze diverse data sets, identify influencing factors, and recommend targeted interventions. The analysis process includes data collection, preprocessing, feature selection, machine learning modeling, and Exploratory data Analysis techniques to understand the dynamics of the Better understanding unemployment among young people. Preliminary results reveal precise predictive patterns and previously unrecognized relationships, providing evidence-based solutions to inform policy and support the economic well-being of younger generations. Further research and validation are needed to increase applicability.

关键词： Technological innovation Renewable energy sources data analysis Machine learning algorithms Wind energy Machine learning Feature extraction

来源：评论

学校读者我要写书评

暂无评论

Kannada Word Detection in Heterogeneous Scene Images 10

Kannada Word Detection in Heterogeneous Scene Images

引用

10th International Conference on Signal Processing and Integrated Networks, SPIN 2023

作者： Shikha, N. Pranav, R. Singh, Nidhi R Umadevi, V. Hussain, Muzammil B.M.S. College of Engineering Dept. of Computer Science Bengaluru India B.M.S. College of Engineering Dept. of Electrical and Electronics Bengaluru India B.M.S. College of Engineering Dept. of Medical Electronics Bengaluru India Tika Data Services Pvt. Ltd. Bengaluru India

ISBN: (纸本)9781665490993

Text extraction from scene images has started gaining a lot of traction in recent years in the computer vision field as its applications is manifold. One of its sub-categories is scene text detection. Factors like complex backgrounds, curvature, orientation, image quality, and various font styles and sizes makes it a difficult task. Moreover, building a single language scene text detector in a multilingual setting, especially Indian scripts, is more challenging in contrast to a general text detector. In this Work We attempt the task of Kannada text localization as Well as Word detection in natural scenes, Which is underexplored in comparison to other Well-known Indian languages. To achieve this, We fine tune a You Only Look Once (YOLOv4) based object detection model on our Kannada scene text images dataset collected. Experiments on local boards and sign images captured on smartphone, showed a mean average precision of 93.5% and per image average detection speed of 30 milli-seconds;pointing towards real time use and integration With mobile applications. © 2023 IEEE.

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

Multiclass Brain Tumor Detection using Deep Transfer Learning

Multiclass Brain Tumor Detection using Deep Transfer Learnin...

引用

MIT Art, Design and Technology School of Computing International Conference (MITADTSoCiCon)

作者： Aishwarya V. Deshmukh M.R. Bendre Dept. of Data Science Engineering Prawara Rural Engineering College Loni India Dept. of Computer Engineering Prawara Rural Engineering College Loni India

ISBN: (数字)9798350362879

ISBN: (纸本)9798350362886

Brain tumors, identified as abnormal growth of cell inside the brain, necessitate accurate categorization for tailored treatment plans. This classification aids in determining the severity of the condition and guiding therapeutic decisions. Iimaging technologies in medical like Magnetic Resonance Imaging (MRI) are instrumental in identifying brain pathologies without emitting ionizing radiation. Deep learning, a subset of AI, enhances brain tumor detection from MRI scans, improving prediction rates. Digital image processing is vital for medical image analysis and diagnosis. Brain tumors manifest in various forms, categorized as benign or malignant based on their tissue characteristics. Convolutional Neural Networks (CNNs), particularly VGG-16, Inception-v3, and ResNet-50 are prominent in brain tumor analysis, segmentation, and classification. These models are compared for their efficacy in identifying and forecasting brain tumor cells. Common brain tumor types include glioma, meningioma, and pituitary tumors, each requiring specific treatment approaches.

关键词： Image segmentation Accuracy Image analysis Magnetic resonance imaging Transfer learning Feature extraction Real-time systems

来源：评论

学校读者我要写书评

暂无评论

Hierarchical randomized smoothing 23

Hierarchical randomized smoothing

引用

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

作者： Yan Scholten Jan Schuchardt Aleksandar Bojchevski Stephan Günnemann Dept. of Computer Science & Munich Data Science Institute Technical University of Munich University of Cologne Germany

Real-world data is complex and often consists of objects that can be decomposed into multiple entities (e.g. images into pixels, graphs into interconnected nodes). Randomized smoothing is a powerful framework for making models provably robust against small changes to their inputs - by guaranteeing robustness of the majority vote when randomly adding noise before classification. Yet, certifying robustness on such complex data via randomized smoothing is challenging when adversaries do not arbitrarily perturb entire objects (e.g. images) but only a subset of their entities (e.g. pixels). As a solution, we introduce hierarchical randomized smoothing: We partially smooth objects by adding random noise only on a randomly selected subset of their entities. By adding noise in a more targeted manner than existing methods we obtain stronger robustness guarantees while maintaining high accuracy. We initialize hierarchical smoothing using different noising distributions, yielding novel robustness certificates for discrete and continuous domains. We experimentally demonstrate the importance of hierarchical smoothing in image and node classification, where it yields superior robustness-accuracy trade-offs. Overall, hierarchical smoothing is an important contribution towards models that are both - certifiably robust to perturbations and accurate. Project page: https://***/daml/hierarchical-smoothing

关键词：

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：