检索结果-内蒙古大学图书馆

2024 Conference on Empirical Methods in Natural Language Processing, EMNLP 2024

作者： Shiri, Fatemeh Guo, Xiao-Yu Far, Mona Golestan Yu, Xin Haffari, Gholamreza Li, Yuan-Fang Department of Data Science & AI Monash University Australia Australian Institute for Machine Learning University of Adelaide Australia School of Electrical Engineering and Computer Science University of Queensland Australia

ISBN: (纸本)9798891761643

Large Multimodal Models (LMMs) have achieved strong performance across a range of vision and language tasks. However, their spatial reasoning capabilities are under-investigated. In this paper, we construct a novel VQA dataset, Spatial-MM, to comprehensively study LMMs' spatial understanding and reasoning capabilities. Our analyses on object-relationship and multi-hop reasoning reveal several important findings. Firstly, bounding boxes and scene graphs, even synthetic ones, can significantly enhance LMMs' spatial reasoning. Secondly, LMMs struggle more with questions posed from the human perspective than the camera perspective about the image. Thirdly, chain of thought (CoT) prompting does not improve model performance on complex multi-hop questions involving spatial relations. Lastly, our perturbation analysis on GQA-spatial reveals that LMMs are much stronger at basic object detection than complex spatial reasoning. We believe our new benchmark dataset and in-depth analyses can spark further research on LMMs spatial reasoning. © 2024 Association for Computational Linguistics.

关键词： Computational linguistics

来源：评论

学校读者我要写书评

暂无评论

Rectifying Demonstration Shortcut in In-Context Learning

Rectifying Demonstration Shortcut in In-Context Learning

引用

2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL 2024

作者： Jang, Joonwon Jang, Sanghwan Kweon, Wonbin Jeon, Minjin Yu, Hwanjo Graduate School of AI POSTECH Korea Republic of Department of Computer Science and Engineering POSTECH Korea Republic of Institute of Artificial Intelligence POSTECH Korea Republic of

ISBN: (纸本)9798891761148

Large language models (LLMs) are able to solve various tasks with only a few demonstrations utilizing their in-context learning (ICL) abilities. However, LLMs often rely on their pre-trained semantic priors of demonstrations rather than on the input-label relationships to proceed with ICL prediction. In this work, we term this phenomenon as the ‘Demonstration Shortcut’. While previous works have primarily focused on improving ICL prediction results for predefined tasks, we aim to rectify the Demonstration Shortcut, thereby enabling the LLM to effectively learn new input-label relationships from demonstrations. To achieve this, we introduce In-Context Calibration, a demonstration-aware calibration method. We evaluate the effectiveness of the proposed method in two settings: (1) the Original ICL Task using the standard label space and (2) the Task Learning setting, where the label space is replaced with semantically unrelated tokens. In both settings, In-Context Calibration demonstrates substantial improvements, with results generalized across three LLM families (OPT, GPT, and Llama2) under various 1 configurations. ©2024 Association for Computational Linguistics.

关键词： Demonstrations

来源：评论

学校读者我要写书评

暂无评论

Parameter-Efficient Federated Cooperative Learning for 3D Object Detection in Autonomous Driving

引用

IEEE Internet of Things Journal 2025年第12期12卷 20314-20325页

作者： Chi, Fangyuan Wang, Yixiao Nasiopoulos, Panos Leung, Victor C.M. University of British Columbia Department of Electrical and Computer Engineering VancouverBCV6T 1Z4 Canada Toronto AI Lab LG Electronics TorontoONM5V 1M3 Canada Shenzhen University College of Computer Science and Software Engineering Shenzhen518060 China

In the rapidly evolving field of autonomous driving, accurately detecting and understanding dynamic environments remains a challenge. Federated learning (FL) offers a promising approach by integrating decentralized models from multiple Connected Autonomous Vehicles (CAVs) to enhance the performance of Deep Learning (DL) based object detection methods. However, traditional FL faces hurdles such as extensive data synchronization requirements, limited data variance, and high communication costs. This paper introduces a federated cooperative learning framework that addresses these challenges by combining data from both CAVs and Roadside Units (RSUs). The framework combines local cooperative perception with global federated learning through a parameter-efficient federated learning adapter (PEFLA) and a lazy communication (LazyComm) strategy, improving DL-based object detection capabilities across diverse driving scenarios while significantly reducing bandwidth requirements. We also present a novel multi-agent-multi-town dataset V2X-Fed, specifically developed to validate the effectiveness of our approach under various conditions. Notably, our framework retains 97.51% of the detection accuracy achieved by full-model federated learning, while utilizing only 1.4% of the bandwidth typically required, demonstrating substantial improvements over conventional federated learning strategies. This study underscores the potential of our tailored approach to substantially enhance autonomous vehicle technologies with minimal resource utilization. © 2014 IEEE.

关键词： Bandwidth

来源：评论

学校读者我要写书评

暂无评论

Technical Analysis of Stock Market Trends using LSTM for Price Prognosis 7

Technical Analysis of Stock Market Trends using LSTM for Pri...

引用

7th International Symposium on Multidisciplinary Studies and Innovative Technologies, ISMSIT 2023

作者： Zade, Nilima Gupta, Kanishk Mutha, Siddhanth Mengshetti, Om Joshi, Gayatri Iyer, R. Kamatchi Lavale Department of Computer Science and Engineering Pune412115 India Universal Ai University Karjat Maharashtra Mumbai India

ISBN: (纸本)9798350342154

This study proposes the utilization of Long Short Term Memory (LSTM) brain networks for specialized analysis of stock exchange patterns and trends to anticipate future stock prices. In this particular study, a comparison is done with three other Machine Learning techniques to show that LSTM is the best implementation. The LSTM model is trained on the previous stock market data, including prices and trading volumes, to learn complex patterns and relationships with stock price trends. The model's accuracy in predicting future stock prices is evaluated, and the results show that LSTM is an effective tool for technical analysis, outperforming traditional time-series forecasting methods. LSTM's ability to capture long term dependencies and unpredictable relationships in the dataset makes it a valuable tool for traders and investors seeking to make informed decisions based on technical analysis of stock market trends. The study demonstrates the potential of LSTM in stock market forecasting, improving investment decision making with accurate price predictions. © 2023 IEEE.

关键词： Long short-term memory

来源：评论

学校读者我要写书评

暂无评论

Deep Learning based Speech Emotion Recognition using Multiple Acoustic Features

Deep Learning based Speech Emotion Recognition using Multipl...

引用

Information Technology, Electronics and Intelligent Communication Systems (ICITEICS), IEEE International Conference on

作者： Shwetkranti Taware Anuradha D. Thakare Department of Computer Engg. Pimpri Chinchwad College of Enggineering Nigadi India Department of Computer Science & Engineering(AI&ML) Pimpri Chinchwad College of Engineering Pune India

ISBN: (数字)9798350382693

ISBN: (纸本)9798350382709

Automatic speech emotion recognition (ASER) is crucial for multiple robotics applications, social media platforms, human-machine interaction systems, auto-chatbots, etc. However, the ASER performance is poor due to inadequate representation of the emotional speech, poor feature distinctiveness, and less generalization capability. This paper presents ASER based on a Deep Convolutional Neural Network (DCNN) that accepts the multiple acoustic features (MAF) to describe emotional speech. Further, it uses a combination of hybrid particle swarm optimization and Archimedes optimization algorithm (PSO-AOA) for selecting the essential features from MAFs. The outcomes of the MAF-PSO-AOA-DCNN are assessed on the BAUM dataset. It provides an overall accuracy of 95.20%, precision of 0.96, recall of 0.94, and F1-score of 0.95, which is superior to traditional techniques.

关键词： Human computer interaction Deep learning Emotion recognition Social networking (online) Speech recognition Acoustics Particle swarm optimization

来源：评论

学校读者我要写书评

暂无评论

Robust Calibration For Improved Weather Prediction Under Distributional Shift

arXiv

引用

arXiv 2024年

作者： Gilda, Sankalp Bhandari, Neel Mak, Wendy Panizza, Andrea ML Collective Department of Computer Science and Engineering RV College of Engineering India

In this paper, we present results on improving out-of-domain weather prediction and uncertainty estimation as part of the Shifts Challenge on Robustness and Uncertainty under Real-World Distributional Shift challenge. We find that by leveraging a mixture of experts in conjunction with an advanced data augmentation technique borrowed from the computer vision domain, in conjunction with robust post-hoc calibration of predictive uncertainties, we can potentially achieve more accurate and better-calibrated results with deep neural networks than with boosted tree models for tabular data. We quantify our predictions using several metrics and propose several future lines of inquiry and experimentation to boost performance. © 2024, CC BY-NC-SA.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

An Effective Trusted and Secure based Clustering and Routing using Improved Bat Optimization Algorithm 2

An Effective Trusted and Secure based Clustering and Routing...

引用

2nd IEEE International Conference on Integrated Circuits and Communication Systems, ICICACS 2024

作者： Al-Attabi, Kassem Aluvala, Srinivas Kodati, Sarangam Anu, D. Swapna, P. The Islamic University Najaf Iraq Sr University School of Computer Science & Artificial Intelligence Warangal India Cvr College of Engineering TS Hyderabad India Nitte Meenakshi Institute of Technology Department of Ai & Ds Bengaluru India Sree Dattha Institute of Engineering and Science Department of Computer Science and Engineering Hyderabad India

ISBN: (纸本)9798350317558

Security measures contribute to the overall longevity of the Wireless Sensor Network (WSN) by preventing attacks that could lead to node compromise or depletion of energy resources. WSNs operate in dynamic environments, necessitating adaptable secure clustering algorithms that can effectively respond to frequent changes in network topology. In this work, we introduce the Improved Bat Optimization algorithm (IBOA) for secure clustering and routing in WSN. We subsequently apply the secure-based trust protocol to transmit data on the secured path. The fitness function is obtained by utilizing the well-defined limitations named energy, delay, security, as well as distance. The proposed method achieves better results and attains an energy consumption of 0.1 (J), an average delay of 0.1 sec, Packet Delivery Ratio (PDR) of 97.5%, and throughput of 80% at the number of nodes of 100 respectively when compared to existing methods such as Dempster-Shafer Theory-based Lion Optimization Algorithm clustering (DST-LOA) and Energy Competent Particle Swarm Optimization depending on the Passive Clustering Mechanism (ECPSO-PCM). © 2024 IEEE.

关键词： Cluster Head Improved Bat Optimization Secured Clustering Trusted path Wireless Sensor Network

来源：评论

学校读者我要写书评

暂无评论

Geological Information Extraction from Satellite Imagery Using Deep Learning

Geological Information Extraction from Satellite Imagery Usi...

引用

International Conference on Computing and Networking Technology (ICCNT)

作者： K. PushpaRani G. Roja R. Anusha C. Dastagiraiah B. Srilatha B. Manjusha Department of Computer Science and Engineering MLR Institute of Technology Hyderabad India Department of CSE(AI & ML) & CSE(DS) GNITS Shaik pet Hyderabad Department of Computer Science and Engineering Anurag University Hyderabad India

ISBN: (数字)9798350370249

ISBN: (纸本)9798350370270

Satellite imagery offers extensive information that can be used for a variety of societal applications, from the number of buildings in a metropolis to the land cover types of a specific area. However, extracting such information is a difficult task. In the past, images were scanned manually by domain experts to extract features like buildings, infrastructure, and vegetation, which would take weeks or even months depending on the image size. To address this challenge, we trained our model using U-Net architecture, a deep learning model, because of its advantages in multiscale feature learning and faster processing. Data augmentation is done on the dataset in order to enhance the amount of data and prevent overfitting artificially. The trained model achieved an F1 score of 92% and an impressive precision score of 95%. This automated method can be scaled to large sets of data, reducing the need for human labeling.

关键词： Deep learning Geology Architecture Buildings Vegetation mapping computer architecture Feature extraction Information retrieval Satellite images Data mining

来源：评论

学校读者我要写书评

暂无评论

Diabetes Prediction Using Machine Learning Techniques: A Case Study from Jordan 25

Diabetes Prediction Using Machine Learning Techniques: A Cas...

引用

25th International Arab Conference on Information Technology, ACIT 2024

作者： Alazaidah, Raed Batyha, Radwan Samara, Ghassan Al Henawi, Esraa Al-Safarini, Maram Y. Samhan, Ahmad A. Elrashidi, Ali Alluwaici, Mo'Ath Al-Omari, Deana It Zarqa University Department of Data Science Ai Faculty Zarqa Jordan Private University Department of Computer Science Faculty of It Applied Science Amman Jordan Zarqa University Department of Computer Science Faculty of It Zarqa Jordan Zarqa University Department of Software Engineering Faculty of It Zarqa Jordan University of Business and Technology Department of Electrical Engineering Jeddah21432 Saudi Arabia

ISBN: (纸本)9798331540012

Diabetes is one of the most common diseases in Jordan. It is the main reason of death among Jordanian adult citizens. Worldwide, 48% of all deaths are due to Diabetes occurred before the age of 70 years. Hence, this research is interested in the early prediction of this disease among Jordanian people. To achieve this main objective, Machine Learning (ml) is utilized through a large number of classification models and considering five well-known evaluation metrics. These classification models have been trained on a primary dataset that has been collected for the purpose of the research. The results revealed that BayesNet and NaiveBayes are the best classification models to handle the task of the early prediction of Diabetes in Jordan © 2024 IEEE.

关键词： Adversarial machine learning

来源：评论

学校读者我要写书评

暂无评论

Enhancing Viewer Engagement: Exploring Generative Long Short Term Memory in Dynamic Ad Customization for Streaming Media 5

Enhancing Viewer Engagement: Exploring Generative Long Short...

引用

5th International Conference on Smart Electronics and Communication, ICOSEC 2024

作者： Ramagundam, Shashishekhar Karne, Niharika Troy University Software Development & Engineering in Content Discovery & Ai Department of computer Science Computer Science Montgomery Campus AL United States C & Rs Platform services DT Amtrak United States University of south alabama Department of electrical engineering Electrical engineering United States

ISBN: (纸本)9798331504403

In Ad customization, several approaches are employed to improve the relevance level of the message for monitoring the user performance according to the prior search history (e.g., retargeting) or by gathering particular information related to the customer (e.g., micro-targeting). Occasionally, the consumers view the ad developed by the other user, in this situation, the ads are viewed as something that can be related by the user. The quantity of information that is gathered openly or secretly can affect the user choice for the content displayed. However, the users remain as recipients of a strategy, that decides the best ad content. Therefore, an exploring Generative Artificial Intelligence-based Dynamic Ad Customization model for enhancing viewer engagement from streaming media is developed. In this paper, the Generative Long Short-Term Memory (GenLSTM) model is developed for examining customized advertising then includes the customization of specific message elements by individuals, variation in the level of ad customization, and in some cases perceived collaboration with others. Here, the customization classification is defined based on the modification in the marketing mix element that is created by the customer proactively as opposed to the company. In regards to customized advertising, 'ad is something that happens with the viewer'. The customization increases based on the sense of task engagement, strength of the perceived customer relationship with the company, attitude toward the ad, and perceptions of company trust/integrity. Finally, the experimental analysis is conducted based on the relationship with the company, company care, company trust, and task engagement to attain the overall impact on ad customization. © 2024 IEEE.

关键词： Sales

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：