检索结果-内蒙古大学图书馆

2023 IEEE PES Innovative Smart Grid Technologies - Asia, ISGT Asia 2023

作者： Patel, Viresh Ghosh, Soumyajit Chakrabarti, Saikat Sharma, Ankush Pannala, Sanjeev Indian Institute of Technology Dept. of Electrical Engineering Uttar Pradesh Kanpur208016 India University of New Brunswick Dept. of Electrical and Computer Engineering FrederictonNB Canada Washington State University School of Electrical Engineering and Computer Science PullmanWA United States

ISBN: (纸本)9798350327748

Locating the fault in the power distribution system is a tedious process that involves manually searching along the line. While some methods rely on complete system parameters or centralized data processing, it has become difficult due to the bidirectional power flow in the active distribution system. This paper proposes a framework consisting of three steps: detecting, classifying, and locating the fault in the active distribution system. The proposed method reduces the search area by using the sending-end sample value of the line current. It can be implemented in a distributed form without knowledge of system parameters using pole-mounted data recording meters and relaying the approximate fault location and type to the control center. In the first step, the fault is detected using an isolation forest. Then, the fault is classified by the support vector machine, and finally, the fault is located using an adaptive weight convolutional neural network (CNN). The CNN weights are modified according to fault type information that uses time-frequency information extracted by a continuous wavelet transform (CWT). The proposed framework is tested using an IEEE 13-node test feeder with a solar photovoltaic system. Different training parameters for the CNN are also tested to analyze the proposed framework © 2023 IEEE.

关键词： Support vector machines

来源：评论

学校读者我要写书评

暂无评论

Hyperspectral Image Classification using Few Shot Domain Adaptation and Active Learning

Hyperspectral Image Classification using Few Shot Domain Ada...

引用

National Conference on Communications (NCC)

作者： Animesh Shree Rakesh Kumar Sanodiya dept. of Computer Science and Engineering Indian Institute of Information Technology Sri City Andhra Pradesh India

ISBN: (数字)9798331536015

ISBN: (纸本)9798331536022

Hyperspectral images (HSIs) are pivotal in remote sensing, providing rich spectral and spatial information for applications such as agriculture, environmental monitoring, and mineral exploration. Unlike traditional RGB images, HSIs capture data across hundreds of contiguous spectral bands, enabling precise material classification based on unique spectral signatures. However, the high dimensionality and the need for pixel-level labeling pose significant challenges, as manual annotation is labor-intensive and time-consuming. To address this, we leverage Domain Adaptation, a transfer learning technique that facilitates knowledge transfer from a well-labeled source domain to a sparsely labeled or unlabeled target domain. In this paper, we focus on few-shot Domain Adaptation for HSI classification, where only a limited number of labeled target samples are available. Our approach integrates Active Learning to iteratively fine-tune a classifier by selecting the most informative target samples, thus reducing domain shift and enhancing model performance. We utilize a framework that generates domain-invariant embeddings by aligning the spectral and spatial features of HSIs from the source and target domains. This method significantly reduces the need for manual labeling while maintaining high classification accuracy. Our experimental results demonstrate the effectiveness of combining few-shot Domain Adaptation with Active Learning, achieving a remarkable 98 % accuracy in hyperspectral image classification with limited labeled data.

关键词： Accuracy Uncertainty Scalability Active learning Transfer learning Manuals Minerals Labeling Hyperspectral imaging Image classification

来源：评论

学校读者我要写书评

暂无评论

Pushing the Limits of Machine Translation with GPT-4 and Marian NMT in Multilingual Communication

Pushing the Limits of Machine Translation with GPT-4 and Mar...

引用

2024 IEEE International Conference on Intelligent Systems and Advanced Applications, ICISAA 2024

作者： Jain, Sourabh Sadim, Mohd Amhia, Hemant Kumar, N. Krishna Choudhury, Shibanjali Rajlakshmi, P.V. GGITS Dept of Computer Science & Engineering M.P. Jabalpur India IES Institute of Technology and Management IES University Department of Computer Science & Engineering M.P. Bhopal462044 India Jabalpur Engineering College Electrical Engineering department MP Jabalpur India Guru Nanak Institutions Technical Campus Dept.of CSE Ibrahimpatam Telangana Hyderabad India Techno College of Engineering Dept. of CSE Tripura Agartala India Kongu Engineering College Department of English TN Erode India

ISBN: (纸本)9798331539948

Machine translation (MT) plays a crucial role in enabling linguistic diversity in online interactions. In this research, we provide a unique method for expanding the capabilities of MT inside a multilingual communication framework by combining GPT-4 with Marian Neural Machine Translation (NMT). We present a complete technique comprising translation quality evaluation, contextual coherence augmentation, and translation delay optimization. The results show that our method is successful in enhancing translation quality, preserving contextual coherence, and decreasing delay. Three algorithms are used in the suggested technique. 1) Contextual Coherence Enhancement, 2) Translation Latency Optimization, and 3) Translation Quality Assessment using BLEU and METEOR ratings. The first method uses BLEU and METEOR scores to quantitatively measure translation quality, guaranteeing linguistic precision. The second method improves conversational fluency by increasing contextual coherence. The third method minimizes the time it takes to translate a sentence, facilitating simultaneous translation and conversation. We evaluated our approach by comparing it to six established MT techniques. According to the outcomes, our strategy is superior to the status quo in many respects. In particular, our approach enhanced translation quality as measured by BLEU and METEOR ratings. It also had a far greater Contextual Coherence Enhancement Factor (CEF) and Coherence Improvement Percentage, indicating that it successfully preserved the rational progressions of conversations. Furthermore, significant improvements in translation latency were also highlighted by the Latency Optimization Ratio (LOR). © 2024 IEEE.

关键词： computer aided language translation

来源：评论

学校读者我要写书评

暂无评论

A Framework for Structural Damage detection system in automobiles for flexible Insurance claim using IOT and Machine Learning

A Framework for Structural Damage detection system in automo...

引用

2022 International Mobile and Embedded technology Conference, MECON 2022

作者： Agarwal, Deepshikha Tripathi, Khushboo Indian Institute of Information Technology Dept. of Information Technology Lucknow India Amity University Dept. of Computer Science and Engineering Haryana Gurgaon India

ISBN: (纸本)9781665420204

Automobiles are required in everyday life to commute from destination to other. But, due to faulty road conditions, heavy traffic and lack of adequate driver skills, crash incidents happen almost every hour of the day. With every vehicle, companies offer insurance which people can claim to get the damaged parts recovered. However, during the recovery, it solely depends on the insurance companies investigating officer to give an adequate claim to the vehicle owner. Due to inadequate visual precipitance, at times, the owner is not given the full insurance claim. The proposed framework offers automated damage detection system by which better and correct claim can be given to such an owner. The proposed framework is based on Machine Learning and IOT approach for transport network scenario. This paper presents the idea of implementing proposed framework to the researchers in future to detect damages and calculate the correct claim to the owner for Insurance claim. © 2022 IEEE.

关键词： Automobiles

来源：评论

学校读者我要写书评

暂无评论

Human Activity Recognition using ShuffleNetV2 Model

Human Activity Recognition using ShuffleNetV2 Model

引用

Intelligent Computing and Control for engineering and Business Systems (ICCEBS)

作者： Shanufa Nazrin P S J.V. Bibal Benifa Anu K. P Aiswarya Vijayakumar dept. of Computer Science St.Joseph's college (university of calicut) Thrissur Kerala dept.of Computer Science and Engineering Indian Institute of Information Technology kottayam Kerala India

HAR is specified as the crucial task within the field of computer vision and artificial intelligence. Main aim of HAR is to observing and labelling different human activity images of distinct data sources by machines such as computer. HAR has applications in various fields. CNN model is an appropriate model for HAR task, due to their ability to automaticaly acquire relevant spatial features, capture hierarchical representations, and handle input of various sizes and types. So it is utilized for developing HAR model. This model can detect objects and also can classify images. The model contains wide ranging catogories. ShuffleNetV2 is a particular architecture within it, mainly for efficient and lightweight deep learning task. This paper develops an intelligent HAR using ShuffleNetV2 model which acquired an accuracy about 69.52 percentage.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Exploring 2D and 3D Face Recognition with FaceNet across Multiple Classifiers

Exploring 2D and 3D Face Recognition with FaceNet across Mul...

引用

Innovative Trends in Information technology (ICITIIT), International Conference on

作者： Priyambada Subudhi Gulshan Kumar Nitigya Joshi Bharath Kumar Gujjari Dept. of Computer Science and Engineering Indian Institute of Information Technology Sri City AP India

ISBN: (数字)9798350386813

ISBN: (纸本)9798350386820

Face recognition technology has witnessed significant advancements in recent years, with researchers exploring novel approaches to enhance accuracy and robustness. This study investigates the performance of FaceNet, a state-of-the-art deep learning model for face recognition, in both 2D and 3D modalities. Additionally, various classifiers are employed to evaluate their impact on the overall recognition accuracy. Both publicly available as well as self-made datasets have been used in this study. The FaceNet model is employed to extract high-dimensional feature vectors from the facial data, leveraging its ability to map faces into a compact Euclidean space. The 2D face recognition system operates on conventional images, while the 3D counterpart processes depth information for a more comprehensive analysis. To evaluate the recognition accuracy, several classifiers are employed, including Support Vector Machines (SVM) and Logistic Regression (LR) model. The classifiers are trained and tested on both 2D and 3D datasets separately. Although, the effectiveness of FaceNet on 2D face recognition has been studied before, this article shows its effectiveness for 3D face recognition as well. Furthermore, the choice of classifier significantly influences the overall recognition accuracy. SVM exhibits robust performance across both modalities, showcasing its versatility in handling complex feature spaces. On the other hand, LR provides a better accuracy in cases where the dataset is equitably distributed.

关键词： Support vector machines Solid modeling Three-dimensional displays Accuracy Webcams Face recognition Feature extraction

来源：评论

学校读者我要写书评

暂无评论

PARAMANU-AYN: Pretrain from scratch or Continual Pretraining of LLMs for Legal Domain Adaptation?

arXiv

引用

arXiv 2024年

作者： Niyogi, Mitodru Bhattacharya, Arnab Gyan AI Research Dept. of Computer Science & Engineering Indian Institute of Technology Kanpur India

In this paper, we present PARAMANU-AYN, a collection of language models based exclusively on legal case documents of India. In particular, the court case documents of the Supreme Court of India (SCI), and text of the Constitution of India and the indian Penal Code are utilized. The novel Auto-Regressive (AR) decoder-only language model is pretrained from scratch at a context size of 8192 on a single A100-PCIE-40GB GPU with an efficient MFU (Model FLOPs Utilization) of 41.35. Our model size is only 97 million parameters. We also trained a tokenizer specialized for legal corpora based on BPE. We evaluated our pretrained legal model on perplexity metric, and did zero-shot evaluation on two tasks: case judgement prediction with explanation and abstractive legal case summarization. Our pretrained legal model outperformed LlaMa-2 7B and Gemini-Pro on zero-shot test accuracy for the legal case judgement prediction with explanation task by close to 2% points, despite being smaller in size by close to 72 times compared to these 7B LLMs. We also reported automatic lexical metrics of zero-shot abstractive summarization generation of our Paramanu-Ayn model across varying summary lengths and compared with decoder-only LLMs at fixed summary length generation (5000 tokens) using ROUGE-1, ROUGE-L, BLEU, METEOR, and BERTScore metrics. Paramanu-Ayn outperformed the models in both BLEU and METEOR metrics by more than 10% percentage points, and also on BERTscore by close to 4% points. Furthermore, we evaluated our model on popular LLM commonsense and mathematical benchmarks for a holistic evaluation despite the model being pretrained exclusively on legal court cases documents. Paramanu-Ayn scored impressively on all the benchmarks. Paramanu-Ayn outperformed Llama-1, Llama-2, and Falcon LLMs on mathematical benchmarks (AGIEVAL-AQuA-RAT, AGIEVAL-SAT-Math) at zero-shot setting. We also instruction-tuned our pretrained model on a set of 10,763 instructions covering various legal tasks such

关键词： Benchmarking

来源：评论

学校读者我要写书评

暂无评论

A Novel Signal Detector based on Approximated Fractional Integrator in Frequency Domain

A Novel Signal Detector based on Approximated Fractional Int...

引用

2023 International Conference on Fractional Differentiation and Its Applications, ICFDA 2023

作者： Kumar, Sumit Sundaram, Gaurav Shourya, Snehan Jha, Rajib Kumar Maharana, Rashmi Ranjan Saini, Garima Indian Institute of Technology Patna Infovision Labs. India Pvt. Ltd. Bangalore India Indian Institute of Technology Patna Dept. of Electrical Engg. Patna India Parala Maharaja Engineering College Dept. of Ete Berhampur India Ims Engineering College Dept. of Computer Science and Engg. Ghaziabad India

ISBN: (纸本)9798350321685

This paper investigates weak signal detection using approximated fractional integrator (AFI). An optimal spectrum sensing scheme based on the criterion of maximization of the deflection ratio coefficient (DR) has been proposed in the frequency domain. Due to the easy availability and less time complexity of fast Fourier transform (FFT) chips in the industry, signal detection is primarily done in the frequency domain. Approximated fractional integrator transforms the data to give the maximum DR because of its non-linear low-pass characteristic. Example and simulation results have been illustrated to observe the performance and properties of the proposed detector. A precise comparison has been discussed with state-of-the-art techniques. © 2023 IEEE.

关键词： Signal detection

来源：评论

学校读者我要写书评

暂无评论

A Novel Encryption Approach with Fractional Discrete Cosine Transform and Cascading Discrete Orthonormal Stockwell Transform

A Novel Encryption Approach with Fractional Discrete Cosine ...

引用

2023 International Conference on Fractional Differentiation and Its Applications, ICFDA 2023

作者： Sundaram, Gaurav Shourya, Snehan Jha, Rajib Kumar Maharana, Rashmi Ranjan Banda, Latha Kumar, Sumit Indian Institute of Technology Patna Dept. of Electrical Engg. Patna India Parala Maharaja Engineering College Dept. of Ete Berhampur India Abes Engineering College Dept. of Computer Science and Engg. Ghaziabad India Indian Institute of Technology Patna Infovision Labs. India Pvt. Ltd. Bangalore India

ISBN: (纸本)9798350321685

This paper proposes a new cryptographic algorithm based on fractional discrete cosine transform (FrDCT) and discrete orthonormal Stockwell (DOST) transform to secure multimedia data. Fractionality makes the cryptosystem more reliable and ensures the security of the data transmitted in the unsecured channel. The proposed method describes the advantage of using FrDCT instead of fractional Fourier transform (FrFT). At the same time, the energy compaction property of DOST guarantees lesser memory and storage requirements. The proposed method has been compared with state-of-the-methods to ensure practicality in different applications. The standard quantitative parameters such as information entropy (≈ 8), NPCR (≈ 99%), UACI (≈ 33%) etc., show its immunity against unauthenticated user(s). © 2023 IEEE.

关键词： Discrete cosine transforms

来源：评论

学校读者我要写书评

暂无评论

Exploring an Inter-Pausal Unit (IPU) based Approach for Indic End-to-End TTS Systems

arXiv

引用

arXiv 2024年

作者： Prakash, Anusha Murthy, Hema A. IITM Bengaluru India Dept of Computer Science & Engineering Indian Institute of Technology Madras Chennai India

Sentences in indian languages are generally longer than those in English. indian languages are also considered to be phrase-based, wherein semantically complete phrases are concatenated to make up sentences. Long utterances lead to poor training of models and result in poor prosody during synthesis. In this work, we explore an inter-pausal unit (IPU) based approach in the end-to-end (E2E) framework, focusing on synthesising conversational-style text. We consider both autoregressive Tacotron2 and non-autoregressive FastSpeech2 architectures in our study and perform experiments with three indian languages, namely, Hindi, Tamil and Telugu. With the IPU-based Tacotron2 approach, we see a reduction in insertion and deletion errors in the synthesised audio, providing an alternative approach to the FastSpeech(2) network in terms of error reduction. The IPU-based approach requires less computational resources and produces prosodically richer synthesis compared to conventional sentence-based systems. © 2024, CC BY.

关键词： Network architecture

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：