检索结果-内蒙古大学图书馆

IEEE Conference on Decision and Control

作者： Romain Cosson Ali Jadbabaie Anuran Makur Amirhossein Reisizadeh Devavrat Shah Inria Paris France Laboratory for Information & Decision Systems MIT Cambridge MA USA Department of Computer Science School of Electrical & Computer Engineering Purdue University West Lafayette IN USA

Several recent empirical studies demonstrate that important machine learning tasks, e.g., training deep neural networks, exhibit low-rank structure, where the loss function varies significantly in only a few directions of the input space. In this paper, we leverage such low-rank structure to reduce the high computational cost of canonical gradient-based methods such as gradient descent (GD). Our proposed Low-Rank Gradient Descent (LRGD) algorithm finds an $\epsilon$ -minimizer of a p-dimensional function by first identifying $r\leq p$ significant directions, and then estimating the true p-dimensional gradient at every iteration by computing directional derivatives only along those $r$ directions. We establish that the “directional oracle complexity” of LRGD for strongly convex objective functions is $\mathrm{O}(r\log(1/\epsilon)+rp)$ . Therefore, when $r\ll p$ , LRGD provides significant improvement over the known complexity of $\mathcal{O}(p\log(1/\epsilon))$ of GD in the strongly convex setting. Furthermore, using real and synthetic data, we empirically find that LRGD provides significant gains over GD when the data has low-rank structure, and in the absence of such structure, LRGD does not degrade performance compared to GD.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Beyond Imagination: Leveraging Generative AI to Enhance Learning Through Story World Analogies

Beyond Imagination: Leveraging Generative AI to Enhance Lear...

引用

Frontiers in Education (FIE) Conference

作者： Dünya Baradari Harry Han Julia Xia Carey Ann Strelecki MIT Media Lab Massachusetts Institute of Technology Cambridge MA Graduate School of Education Harvard University Cambridge MA Electrical Engineering and Computer Science Massachusetts Institute of Technology Cambridge MA

ISBN: (数字)9798350351507

ISBN: (纸本)9798350363067

This innovative practice full paper describes ConceptualTales, a conversational AI that explains STEM and social science concepts using analogies from popular story worlds and Socratic reasoning. The disconnect between conventional teaching strategies and student engagement is a persistent challenge in educational systems, particularly STEM fields. Traditional methods often fail to resonate with students, rendering the learning process monotonous and detached from their personal interests. Concurrently, students are enthusiastic about and dedicated to fictional worlds such as Marvel, Harry Potter, and Disney. This observation forms the basis for our innovative practice: integrating these beloved narratives into educational content through generative AI. ConceptualTales was tried with middle and high school students in the USA and China and received overwhelmingly positive feedback. Our system combines fiction-inspired learning, analogical reasoning, and Socratic questions, to make educational content personal and interesting to students.

关键词： Conversational artificial intelligence Generative AI Social sciences Education Rendering (computer graphics) Cognition STEM

来源：评论

学校读者我要写书评

暂无评论

Current Implicit Policies May Not Eradicate COVID-19

arXiv

引用

arXiv 2022年

作者： Jadbabaie, Ali Sarker, Arnab Shah, Devavrat Institute for Data Systems and Society MIT United States Department of Civil and Environmental Engineering MIT United States Department of Electrical Engineering and Computer Science MIT United States

Successful predictive modeling of epidemics requires an understanding of the implicit feedback control strategies which are implemented by populations to modulate the spread of contagion. While this task of capturing endogenous behavior can be achieved through intricate modeling assumptions, we find that a population's reaction to case counts can be described through a second order affine dynamical system with linear control which fits well to the data across different regions and times throughout the COVID-19 pandemic. The model fits the data well both in and out of sample across the 50 states of the United States, with comparable R2 scores to state of the art ensemble predictions. In contrast to recent models of epidemics, rather than assuming that individuals directly control the contact rate which governs the spread of disease, we assume that individuals control the rate at which they vary their number of interactions, i.e. they control the derivative of the contact rate. We propose an implicit feedback law for this control input and verify that it correlates with policies taken throughout the pandemic. A key takeaway of the dynamical model is that the "stable" point of case counts is non-zero, i.e. COVID-19 will not be eradicated under the current collection of policies and strategies, and additional policies are needed to fully eradicate it quickly. Hence, we suggest alternative implicit policies which focus on making interventions (such as vaccinations and mobility restrictions) a function of cumulative case counts, for which our results suggest a better possibility of eradicating COVID-19. Copyright © 2022, The Authors. All rights reserved.

关键词： Linear control systems

来源：评论

学校读者我要写书评

暂无评论

IoT and wireless sensor network -Based Three- Tier Architecture for Continuous Cardiac Health Monitoring and Alert System

IoT and wireless sensor network -Based Three- Tier Architect...

引用

Expert Clouds and Applications (ICOECA), International Conference on

作者： Swapnil Gulabrao Gorde Archana Singh M Tamilselvi G Ravivarman Rajendiran M Ramya Maranan MIT College of Management MIT Art Design & Technology University Pune Private University Maharashtra India Department of Computer Science and Engineering Roever Engineering College Perambalur Tamil Nadu India Department of Electrical and Electronics Engineering Karpagam Academy of Higher Education Coimbatore Tamil Nadu India Department of Computer Science and Engineering Panimalar Engineering College Poonamallee Chennai Tamil Nadu India Department of Research and Innovation SIMATS Saveetha School of Engineering Chennai Tamil Nadu India

ISBN: (数字)9798350385793

ISBN: (纸本)9798350385809

This study suggests a powerful Three- Tier Architecture that combines machine learning models, Internet of Things (IoT) devices, and sophisticated sensors to provide continuous cardiac health monitoring. A range of physiological data is provided by the sensors in the Data Collection Tier, which measure blood pressure, temperature, acceleration, and electrocardiogram (ECG). In the Data Processing Tier, a local server applies complex algorithms to efficiently assess the gathered data and provide insightful outcomes. By providing easy access, authentication, and alarm production, the Application Tier acts as the interface and guarantees efficient contact with healthcare providers. The unique aspect of the system is the integration of machine learning models, which were trained on recorded data for predictive analysis. k-Nearest Neighbors (KNN), Artificial Neural Networks (ANN), Decision Trees (DT), and Recurrent Neural Networks (RNN) are some examples of these models. Impressive accuracy levels are seen in the results, with ANN leading at 96.44% The measures of accuracy, precision, recall, and F1 score all demonstrate the algorithms' capacity to anticipate possible health risks, which supports the models' effectiveness. This proactive paradigm in healthcare is made possible by the comprehensive approach of continuous heart health monitoring, which enables prompt interventions and customized treatment strategies. This study increases our knowledge of the dynamics of cardiovascular health and the rapidly growing field of health care technology by demonstrating the synergy of sensor-based monitoring, IoT connectivity, and machine learning analytics in a unified architecture. In the end, patient care is improved.

关键词： Temperature sensors Temperature measurement Machine learning algorithms Accuracy Recurrent neural networks Artificial neural networks Sensors

来源：评论

学校读者我要写书评

暂无评论

Machine Learning for Risk Prediction: Does One Size Really Fit All?

JACC: Advances

引用

JACC: Advances 2023年第7期2卷 100552-100552页

作者： Stultz, Collin M. Department of Electrical Engineering and Computer Science and Institute for Medical Engineering & Science MIT Cambridge MA United States Division of Cardiology MGH Boston MA United States

Corresponding Author

关键词： decision trees machine learning risk stratification

来源：评论

学校读者我要写书评

暂无评论

Combinatorial NLTS From the Overlap Gap Property

arXiv

引用

arXiv 2023年

作者： Anschuetz, Eric R. Gamarnik, David Kiani, Bobak MIT Center for Theoretical Physics CambridgeMA02139 United States Sloan School of Management MIT CambridgeMA02139 United States Department of Electrical Engineering and Computer Science MIT CambridgeMA02139 United States

In an important recent development, Anshu, Breuckmann, and Nirkhe [ABN23] resolved positively the so-called No Low-Energy Trivial State (NLTS) conjecture by Freedman and Hastings [FH14]. The conjecture postulated the existence of linear-size local Hamiltonians on n qubit systems for which no near-ground state can be prepared by a shallow (sublogarithmic depth) circuit. The construction in [ABN23] is based on recently developed good quantum codes. Earlier results in this direction included the constructions of the so-called Combinatorial NLTS—a weaker version of NLTS—where a state is defined to have low energy if it violates at most a vanishing fraction of the Hamiltonian terms [EH17],[AB22]. These constructions were also based on codes. In this paper we provide a "non-code" construction of a class of Hamiltonians satisfying the Combinatorial NLTS. The construction is inspired by one in [AB22], but our proof uses the complex solution space geometry of random K-SAT instead of properties of codes. Specifically, it is known that above a certain clause-to-variables density the set of satisfying assignments of random K-SAT exhibits the overlap gap property, which implies that it can be partitioned into exponentially many clusters each constituting at most an exponentially small fraction of the total set of satisfying solutions [ACORT11]. We establish a certain robust version of this clustering property for the space of near-satisfying assignments and show that for our constructed Hamiltonians every combinatorial near-ground state induces a near-uniform distribution supported by this set. Standard arguments then are used to show that such distributions cannot be prepared by quantum circuits with depth o(log n). Unfortunately, our Hamiltonian is only local on average due to fluctuations in the numbers of clauses in which variables are involved. On the positive side, since the clustering property is exhibited by many random structures—including proper coloring, maximum cut,

关键词： Hamiltonians

来源：评论

学校读者我要写书评

暂无评论

Cloud-based Machine Learning Adoption Model for Higher Education Institutions

Cloud-based Machine Learning Adoption Model for Higher Educa...

引用

2024 International Conference on Artificial Intelligence and Emerging Technology, Global AI Summit 2024

作者： Kiros, Atakilti Brhanu Salau, Ayodeji Olalekan Vats, Satvik Kaushik, Keshav Tin, Ting Tin Omeje, Crescent Onyebuchi Mekelle University Mekelle Institute of Technology Mit Mekelle Ethiopia Afe Babalola University Department of Electrical/Electronics and Computer Engineering Ado-Ekiti Nigeria Department of Cse Graphic Era Hill University Uttarakhand Dehradun India Graphic Era Deemed to be University Uttarakhand Dehradun248007 India Amity School of Engineering and Technology Amity University Mohali Punjab India Inti International University Faculty of Data Science and Information Technology Negeri Sembilan Malaysia University of Port Harcourt Department of Electrical/Electronic Engineering Rivers State Choba500004 Nigeria

ISBN: (纸本)9798350379716

As a new concept, cloud computing has attracted the IT Initiations attention especially the Mobile learning (M-learning). At present, there Higher Education Institutions (HEI's) have struggle with limited IT budgets, difficulties in obtaining and managing infrastructure licenses, and the complexities of software and hardware management. The coming of cloud computing, all of the above challenges was solved. The migration to cloud based M-learning in HEI's is a significant stage in terms of online education, especially during the COVID-19 pandemic. Cloud based M-learning approach has the potential to be a powerful and rapid solution to the many challenges faced by universities in this unprecedented time. This study starts with some brief concepts of Cloud Computing and M-learning for HEI's, mentioning for the significance of the results obtained so far. Additionally, it starting point as strategy for HEI's to adopt Cloud based M-learning is provided, by comparisons of existing models and select DOI model to propose cloud based adoption model. But, the DOI model has its own weaknesses;for example, it does not take into account the localization as Innovation determinants. This study introduces localization phase for the optimal performance of the adopted model. The adoption model was evaluating the importance of the proposed model by IT domain experts and it initiates to be both useful and suitable for a successful cloud based M-learning adoption Ethiopian perspective. Therefore, the research results obtained are promising and funding the use of Cloud based M-learning solutions for Ethiopia HEI's by improving knowledge and skill in this field provides a useful strategy for the HEI's structure. © 2024 IEEE.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Identifying Cyber Extremism Sentiments using ROBERTA

Identifying Cyber Extremism Sentiments using ROBERTA

引用

Intelligent and Innovative Technologies in Computing, electrical and Electronics (IITCEE), International Conference

作者： Sudhanshu Patel K. Raja J. Shiny Duela Thomas M Chen mithileysh Sathiyanarayanan Department of Computer Science and Engineering SRM Institute of Science and Technology Chennai India Department of Electrical and Electronic Engineering School of Science and Technology City University of London Research & Innovation MIT Square London City University of London

It is critical to identify and mitigate cyber extremist sentiments in order to prevent terrorist acts and maintain national security. The use of online platforms to promote extremist ideologies and incite violence is referred to as cyber extremism. To detect and classify extremist content on the internet, machine learning models have been used, making the process more efficient and scalable. To identify cyber extremism sentiments using machine learning models, the model is typically trained on a large dataset of extremist and non-extremist materials. The model then learns to recognize patterns and features associated with extremism, such as the use of specific words and phrases, as well as image and video-based features like facial expressions and gestures. We use ROBERTA for this purpose in this study and compare its results to those of BERT and ALBERT. The use of machine learning to detect cyber extremism sentiments can improve law enforcement and other organizations' ability to detect and respond to potential threats, ultimately helping to keep communities safe.

关键词： Training Ethics Image recognition Law enforcement Natural languages Machine learning Organizations

来源：评论

学校读者我要写书评

暂无评论

electrically Programmable Non-Volatile Silicon Photonic Content Addressable Memory (CAM) cell

Electrically Programmable Non-Volatile Silicon Photonic Cont...

引用

CLEO: science and Innovations in CLEO 2024, CLEO: S and I 2024 - Part of Conference on Lasers and Electro-Optics

作者： Lian, Chuanyu Sun, Hongyi Huang, Yi-Siou Vitale, Steven A. Hu, Juejun Takeuchi, Ichiro Youngblood, Nathan Vagionas, Christos Ríos Ocampo, Carlos A. Department of Materials Science & Engineering University of Maryland College ParkMD20742 United States Institute for Research in Electronics and Applied Physics University of Maryland College ParkMD20742 United States Advanced Materials and Microsystems Group MIT Lincoln Laboratory LexingtonMA02139 United States Department of Materials Science & Engineering MIT CambridgeMA02139 United States Quantum Materials Center Department of Physics University of Maryland College ParkMD20742 United States Electrical & Computer Engineering Department The University of Pittsburgh PittsburghPA15213 United States Department of Informatics Center for Interdisciplinary Research and Innovation Aristotle University of Thessaloniki Thessaloniki Greece

We experimentally demonstrate the first electrically programmable, non-volatile silicon photonic content addressable memory cell using Sb2Se3 phase change material on microring resonators, opening the path for light-b... 详细信息

关键词： Silicon photonics

来源：评论

学校读者我要写书评

暂无评论

Position: a roadmap to pluralistic alignment 24

Position: a roadmap to pluralistic alignment

引用

Proceedings of the 41st International Conference on Machine Learning

作者： Taylor Sorensen Jared Moore Jillian Fisher mitchell Gordon Niloofar Mireshghallah Christopher Michael Rytting Andre Ye Liwei Jiang Ximing Lu Nouha Dziri Tim Althoff Yejin Choi Department of Computer Science University of Washington Seattle Washington Department of Computer Science Stanford University Stanford California Department of Computer Science and Department of Statistics University of Washington Seattle Washington Department of Computer Science University of Washington Seattle Washington and Department of Electrical Engineering and Computer Science MIT Cambridge Massachusetts Department of Computer Science University of Washington Seattle Washington and Allen Institute for Artificial Intelligence Seattle Washington Allen Institute for Artificial Intelligence Seattle Washington

With increased power and prevalence of AI systems, it is ever more critical that AI systems are designed to serve all, i.e., people with diverse values and perspectives. However, aligning models to serve pluralistic human values remains an open research question. In this piece, we propose a roadmap to pluralistic alignment, specifically using large language models as a test bed. We identify and formalize three possible ways to define and operationalize pluralism in AI systems: 1) Overton pluralistic models that present a spectrum of reasonable responses; 2) Steerably pluralistic models that can steer to reflect certain perspectives; and 3) Distributionally pluralistic models that are well-calibrated to a given population in distribution. We also formalize and discuss three possible classes of pluralistic benchmarks: 1) Multi-objective benchmarks, 2) Tradeoff steerable benchmarks that incentivize models to steer to arbitrary trade-offs, and 3) Jurypluralistic benchmarks that explicitly model diverse human ratings. We use this framework to argue that current alignment techniques may be fundamentally limited for pluralistic AI; indeed, we highlight empirical evidence, both from our own experiments and from other work, that standard alignment procedures might reduce distributional pluralism in models, motivating the need for further research on pluralistic alignment.

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：