检索结果-内蒙古大学图书馆

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Liu, Che Ouyang, Cheng Cheng, Sibo Shah, Anand Bai, Wenjia Arcucci, Rossella Department of Earth Science and Engineering Imperial College London United Kingdom Data Science Institute Imperial College London United Kingdom Department of Computing Imperial College London United Kingdom Department of Brain Sciences Imperial College London United Kingdom Department of Infectious Disease Epidemiology Imperial College London United Kingdom Royal Brompton and Harefield Hospitals United Kingdom Department of Engineering Science University of Oxford Oxford United Kingdom Institute of Clinical Sciences Imperial College London United Kingdom CEREA École des Ponts and EDF R&D Île-de-France France

Medical imaging tasks require an understanding of subtle and localized visual features due to the inherently detailed and area-specific nature of pathological patterns, which are crucial for clinical diagnosis. Although recent advances in medical vision-language pre-training (VLP) enable models to learn clinically relevant visual features by leveraging both medical images and their associated radiology reports, current medical VLP methods primarily focus on aligning images with entire reports. This focus hinders the learning of dense (pixel-level) visual features and is suboptimal for dense prediction tasks (e.g., medical image segmentation). To address this challenge, we propose a novel medical VLP framework, named Global to Dense level representation learning (G2D), which aims to learn global and dense visual features simultaneously using only image-text pairs without extra annotations. In particular, G2D designs a Pseudo Segmentation (PS) task, which enables the model to learn dense visual features during VLP. Notably, generating PS masks can be performed on the fly during VLP, which does not incur extra trainable parameters. With this simple yet effective idea, G2D achieves superior performance across 5 medical imaging tasks and 25 diseases. Particularly, in the segmentation task which requires dense visual features, G2D surpasses existing models even with just 1% of the training data for finetuning, compared to 100% used by other models. The code can be found in https://***/cheliucomputation/G2D-NeurIPS24/tree/main. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Automated Detection of Acute Appendicitis from Contrast-enhanced CT Scans using Generative Adversarial Networks for Augmented Classification in Acute Abdomen Cases

Automated Detection of Acute Appendicitis from Contrast-enha...

引用

Smart Technologies, Communication and Robotics (STCR)

作者： J. R. Marshal Done J. Srija J. Jerlin Rajan Division of Data Science and Cyber Security Karunya Institute of Technology and Sciences Coimbatore Tamil Nadu India Karunya School of Management Karunya Institute of Technology and Sciences Coimbatore Tamil Nadu India

ISBN: (数字)9798350357530

ISBN: (纸本)9798350357547

The research introduces an innovative method to detect acute appendicitis by uniting symptom-based forecasting with state-of-the-art image analysis techniques. The system employs an FNN for syndrome evaluation which interacts with a 3D ResNet model for analyzing CT scan contrast images with the objective to enhance diagnostic speed while increasing accuracy. Users supply patient symptoms to an FNN model that calculates appendicitis risk probabilities. User needs to upload CT scan files when the system detects probabilities exceeding the designated threshold. The system uses 3D ResNet to examine the uploaded images. Users benefit from the MERN stack-based website because it delivers an intuitive experience for both symptom entry and CT scan file upload capabilities. The machine learning models that run for real-time predictions operate on FastAPI infrastructure. This two-level diagnostic process quickly delivers precise appendicitis diagnosis that assists health care professionals in their medical choices. A forthcoming enhancement plan includes the addition of components for patient management as well as appointment scheduling while adding extra predictive models to enhance system strength.

关键词： Solid modeling Analytical models Three-dimensional displays Accuracy Computed tomography Predictive models Generative adversarial networks Real-time systems Medical diagnostic imaging System implementation

来源：评论

学校读者我要写书评

暂无评论

Artificial Intelligence and data science for Social Good: The Case for the SDGs

Artificial Intelligence and Data Science for Social Good: Th...

引用

2022 IEEE/WIC/ACM International Joint Conference on Web Intelligence and Intelligent Agent Technology (WI-IAT)

作者： Maria Fasli Business and Local Government Data Research Centre School of Computer Science and Electronic Engineering Institute for Analytics and Data Science University of Essex Colchester United Kingdom

Technologies have always been reshaping our economic, social and cultural environments but more recently we have seen the speed of change and impact of ICT technologies in particular rapidly increasing. These phenomenal changes are fuelled by the proliferation of data and the advances in computational methods. Though Artificial Intelligence as a (sub)discipline has been around for over 60 years, it is only in the last decade or so that we have seen very significant improvements in techniques such as machine learning. The potential of Artificial Intelligence to transform multiple aspects of our lives is huge, but there is an equally huge challenge in ensuring that the benefits of these technologies reach all.

关键词： Economics Transforms Machine learning data science Intelligent agents Cultural differences

来源：评论

学校读者我要写书评

暂无评论

Predictive Modeling and Machine Learning Insights into Multi-Drug Resistant Tuberculosis Treatment Outcomes

Predictive Modeling and Machine Learning Insights into Multi...

引用

IEEE Symposium on Systems and Information Engineering Design, SIEDS

作者： Lucille Shichman Kayla Lee Farah Turkistani Adam Tashman Donald Brown Scott Heysell Peter Mbelele Stellah Mpagama Kennedy Ngowi University of Virginia School of Data Science United States University of Virginia School of Medicine United States Kibong’oto Infectious Disease Hospital Tanzania Kilimanjaro Clinical Research Institute Tanzania

ISBN: (数字)9798331535759

ISBN: (纸本)9798331535766

Tuberculosis (TB) remains a major global health burden, particularly in resource-prevented settings. Predictive models may improve patient outcomes by identifying individuals at higher risk of treatment failure, but many existing models are trained on data from high-income countries and may not generalize well to low- and middle-income contexts. In this study, we analyzed two datasets: a longitudinal, multi-site cohort, termed the International Collaborations in Infectious Disease Research (ICIDR), with data from Tanzania, Bangladesh, and Siberia and a cross-sectional dataset from Kibong’oto Infectious Disease Hospital (KIDH) in Tanzania. We applied logistic regression, random forest, and XGBoost models to predict outcomes in multidrug-resistant TB (MDR-TB). For the KIDH dataset, patients were stratified by treatment regimen (bedaquiline, short-term injectable, and long-term injectable) to capture regimen-specific predictors. Models trained on the ICIDR dataset showed limited predictive power, likely due to data sparsity and irregular follow-up. In contrast, stratified XGBoost models trained on the KIDH dataset performed better, with F1 scores of 0.766, 0.667, and 0.787 for the respective regimens. BMI consistently emerged as a top predictor, underscoring the role of nutrition in TB recovery. These results suggest that stratifying predictive models by treatment regimen improves both performance and interpretability. However, small sample sizes and inconsistent data collection remain key limitations. Future studies should prioritize standardized, prospective data collection and interdisciplinary collaboration to develop robust, contextually relevant predictive tools.

关键词： Resistance Logistic regression Tuberculosis Infectious diseases Hospitals Predictive models data collection data models International collaboration Random forests

来源：评论

学校读者我要写书评

暂无评论

Deep Learning for Emoji Generation: From Faces and Gestures to Digital Emotion

Deep Learning for Emoji Generation: From Faces and Gestures ...

引用

Integration of Emerging Technologies for the Digital World (ICIETDW), International Conference on

作者： R. Nareshkumar K. Nimala M. K. Mohamed Faizal P. Sasikumar S. Karthigeyan Srikanth Cherukuvada Department of Networking and Communications School of Computing College of Engineering and Technology SRM Institute of Science and Technology Tamilnadu India Department of Artificial Intelligence and Data Science K. Ramakrishnan College of Engineering Samayapuram Tamil Nadu India Department of AIML Sphoorthy Engineering College Hyderabad Tamilnadu India Department of Computer Science and Engineering Amrita School of Computing Amrita Vishwa Vidyapeetham Chennai India Department of Computer Science and Engineering Koneru Lakshmaiah Education Foundation Hyderabad Telangana India

ISBN: (数字)9798331541170

ISBN: (纸本)9798331541187

The objective of this project is to create a reliable and effective system for identifying facial expressions and hand movements and translating them into suitable emojis. Emotions are vital in human communication, and precisely understanding them may significantly improve user experiences in several applications, such as social media, virtual assistants, and human-computer interaction. Through the creation of an extensive dataset and the use of a robust CNN, the system is capable of precisely categorizing facial expressions, even when faced with different lighting conditions and a wide range of facial characteristics. The suggested methodology entails the acquisition of live video data, which is then subjected to preprocessing to enhance the extraction of relevant features. Subsequently, CNN models are used to accurately recognize and interpret facial emotions and movements. These acknowledged phrases are then converted into equivalent emojis, offering an instinctive and smooth method to communicate emotions via digital communication. Initial findings indicate that the system attains a notable level of precision in identifying a diverse array of facial emotions, showcasing its capacity for practical use in real-life scenarios. This study enhances the field of human-computer interaction by facilitating a more authentic and articulate form of communication via emojis. It effectively narrows the divide between human emotions and digital interfaces. Subsequent efforts will concentrate on augmenting the system's functionalities to include a wider array of motions and improving its efficiency in more demanding settings.

关键词： Human computer interaction Hands Deep learning Emotion recognition Social networking (online) Virtual assistants Feature extraction Digital communication Real-time systems Emojis

来源：评论

学校读者我要写书评

暂无评论

BWArea Model: Learning World Model, Inverse Dynamics, and Policy for Controllable Language Generation

arXiv

引用

arXiv 2024年

作者： Jia, Chengxing Wang, Pengyuan Li, Ziniu Li, Yi-Chen Zhang, Zhilong Tang, Nan Yu, Yang National Key Laboratory for Novel Software Technology Nanjing University China School of Artificial Intelligence Nanjing University China School of Data Science The Chinese University of Hong Kong Shenzhen China Shenzhen Research Institute of Big Data China Polixir.ai

Large language models (LLMs) have catalyzed a paradigm shift in natural language processing, yet their limited controllability poses a significant challenge for downstream applications. We aim to address this by drawing inspiration from the neural mechanisms of the human brain, specifically Broca’s and Wernicke’s areas, which are crucial for language generation and comprehension, respectively. In particular, Broca’s area receives cognitive decision signals from Wernicke’s area, treating the language generation as an intricate decision-making process, which differs from the fully auto-regressive language generation of existing LLMs. In a similar vein, our proposed system, the BWArea model, conceptualizes language generation as a decision-making task. This model has three components: a language world model, an inverse dynamics model, and a cognitive policy. Like Wernicke’s area, the inverse dynamics model is designed to deduce the underlying cognitive intentions, or latent actions, behind each token. The BWArea model is amenable to both pre-training and fine-tuning like existing LLMs. With 30B clean pre-training tokens, we have trained a BWArea model, which achieves competitive performance with LLMs of equal size (1B parameters). Unlike fully auto-regressive LLMs, its pre-training performance does not degenerate if dirty data unintentionally appears. This shows the advantage of a decomposed structure of BWArea model in reducing efforts in laborious data selection and labeling. Finally, we reveal that the BWArea model offers enhanced controllability via fine-tuning the cognitive policy with downstream reward metrics, thereby facilitating alignment with greater simplicity. On 9 out of 10 tasks from two suites, TextWorld and BigBench Hard, our method shows superior performance to auto-regressive LLMs. © 2024, CC BY-NC-ND.

关键词： Decision making

来源：评论

学校读者我要写书评

暂无评论

E1 TTS: Simple and Fast Non-Autoregressive TTS

arXiv

引用

arXiv 2024年

作者： Liu, Zhijun Wang, Shuai Zhu, Pengcheng Bi, Mengxiao Li, Haizhou School of Data Science The Chinese University of Hong Kong Guangdong Shenzhen China Shenzhen Research Institute of Big Data The Chinese University of Hong Kong Guangdong Shenzhen China Fuxi AI Lab NetEase Inc. Hangzhou China

This paper introduces Easy One-Step Text-to-Speech (E1 TTS), an efficient non-autoregressive zero-shot text-to-speech system based on denoising diffusion pretraining and distribution matching distillation. The training of E1 TTS is straightforward;it does not require explicit monotonic alignment between the text and audio pairs. The inference of E1 TTS is efficient, requiring only one neural network evaluation for each utterance. Despite its sampling efficiency, E1 TTS achieves naturalness and speaker similarity comparable to various strong baseline models. Audio samples are available at ***. © 2024, CC BY.

关键词： Neural networks

来源：评论

学校读者我要写书评

暂无评论

Integrating I4.0 Knowledge Graphs with Large Language Models Beyond SPARQL Endpoints 1

Integrating I4.0 Knowledge Graphs with Large Language Models...

引用

Joint of the 1st Software Lifecycle Management for Knowledge Graphs Workshop and the 3rd International Workshop on Semantic Industrial Information Modelling, SOFLIM2KG-SEMIIM 2024

作者： Wahid, Abdul Yahya, Muhammad Zaman, Farooq Zhou, Baifan Breslin, John G. Intizar, Muhammad Ali Kharlamov, Evgeny Data Science Institute University of Galway IDA Business Park Lower Dangan GalwayH91 AEX4 Ireland Valeo Vision Systems Tuam Galways Ireland Information Technology University Lahore Pakistan Department of Computer Science Oslo Metropolitan University Oslo Norway School of Electronic Engineering Dublin City University Dublin Ireland Bosch Center for AI Renningen Germany

Industry 4.0 (I4.0) knowledge graphs are a common way to represent industrial information models. Conventional SPARQL querying systems require the users to be familiar with the data schema and SPARQL syntax. However, this is often very difficult for many users in industrial production, who have mostly an engineering background, instead of a semantic web. Recent developments in large language models (LLMs) make it possible for non-semantic experts to use natural language to query knowledge graphs (KG). In this work, we present a framework and preliminary results of integrating Industry 4.0 KGs with LLMs to improve how data is represented, reasoned, and processed in manufacturing contexts, facilitating user interaction with KGs and contributing to operational efficiency. Our technique enhances Language Models (LLMs) by utilising the semantic complexity and interdependence of Knowledge Graphs (KGs). This allows us to incorporate domain-specific knowledge. We used the FAISS library and LLaMA2 to optimise the storage and retrieval of vectors, which improved the system’s performance and scalability. This integration allows for advanced fault detection, proactive maintenance, and process optimisation, resulting in decreased periods of inactivity and improved productivity. We introduce the framework’s architecture, implementation strategy, and possible advantages while also discussing the difficulties associated with data integration and scalability. The results of our study show that the integration of KG-LLM surpasses traditional approaches in terms of operational efficiency, as evidenced by enhanced fault detection, proactive maintenance, and process optimisation, thereby opening up possibilities for the advancement of more intelligent and resilient production systems. © 2022 Copyright for this paper by its authors.

关键词： Knowledge graph

来源：评论

学校读者我要写书评

暂无评论

Foundational data science Training for Health Equity Researchers at Minority Serving Institutions: A SoRDS Event 11

Foundational Data Science Training for Health Equity Researc...

引用

11th IEEE International Conference on Healthcare Informatics, ICHI 2023

作者： Quick, Robert Córdoba, Marcela Alfaro Diggs, Stephen Cobe, Raphael Bezuidenhout, Louise Shannahan, Hugh Peterson, Bianca Indiana University Pervasive Technology Institute BloomingtonIN United States University of California Santa Cruz Dept of Statistics Santa Cruz United States California Digital Library OaklandCA United States São Paulo State University Center for Scientific Computing São Paulo Brazil Data Archiving and Networked Services The Hague Netherlands Royal Holloway University of London Department of Computer Science Egham United Kingdom Fathom Data Potchefstroom South Africa

ISBN: (纸本)9798350302639

Contemporary research, particularly when addressing the most significant transdisciplinary research challenges, cannot effectively be done without a range of skills relating to data management, data analysis, and cyberinfrastructure (CI). These data and CI skills are common to all disciplines that conduct data-centric research. Research data science acts as a vital component of the scientific process. In a grassroots attempt to address this gap, the COdata-RDA schools of Research data science (SoRDS) was founded in 2016 to provide instruction on foundational data science and open research concepts to early career researchers in low and middle-income countries. This partnership between international collaborators has since 2016 provided this training to over 1000 early career researchers in 24 events in 10 countries worldwide. The most recent event was held at Georgia institute of Technology and focused on health equity and included researchers from minority-serving institutions in the southeast United States. This paper covers the background of the SoRDS project along with organization and curriculum details. It also covers the transition of the events from a non-domain-centric curriculum to spotlighting biological and social health equity data and what we learned to make future health-related events more engaging and valuable to the attendees. It also looks toward future events that will serve international students studying health informatics and other data-centric disciplines. © 2023 IEEE.

关键词： Open data

来源：评论

学校读者我要写书评

暂无评论

Towards Model Resistant to Transferable Adversarial Examples via Trigger Activation

arXiv

引用

arXiv 2025年

作者： Yu, Yi Xia, Song Lin, Xun Kong, Chenqi Yang, Wenhan Lu, Shijian Tan, Yap-Peng Kot, Alex C. Lab Interdisciplinary Graduate Programme Nanyang Technological University Singapore PengCheng Laboratory Shenzhen China School of Electrical and Electronic Engineering Nanyang Technological University Singapore School of Computer Science and Engineering Beihang University Beijing China College of Computing and Data Science Nanyang Technological University Singapore

Adversarial examples, characterized by imperceptible perturbations, pose significant threats to deep neural networks by misleading their predictions. A critical aspect of these examples is their transferability, allowing them to deceive unseen models in black-box scenarios. Despite the widespread exploration of defense methods, including those on transferability, they show limitations: inefficient deployment, ineffective defense, and degraded performance on clean images. In this work, we introduce a novel training paradigm aimed at enhancing robustness against transferable adversarial examples (TAEs) in a more efficient and effective way. We propose a model that exhibits random guessing behavior when presented with clean data x as input, and generates accurate predictions when with triggered data x + τ. Importantly, the trigger τ remains constant for all data instances. We refer to these models as models with trigger activation. We are surprised to find that these models exhibit certain robustness against TAEs. Through the consideration of first-order gradients, we provide a theoretical analysis of this robustness. Moreover, through the joint optimization of the learnable trigger and the model, we achieve improved robustness to transferable attacks. Extensive experiments conducted across diverse datasets, evaluating a variety of attacking methods, underscore the effectiveness and superiority of our approach. Copyright © 2025, The Authors. All rights reserved.

关键词： Deep neural networks

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：