检索结果-内蒙古大学图书馆

SWING-BY DYNAMICS IN CONCEPT LEARNING AND COMPOSITIONAL GENERALIZATION

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Yang, Yongyi Park, Core Francisco Lubana, Ekdeep Singh Okawa, Maya Hu, Wei Tanaka, Hidenori CBS-NTT Physics of Intelligence Program Harvard University CambridgeMA United States Physics & Informatics Laboratories NTT Research Inc. SunnyvaleCA United States Computer Science and Engineering University of Michigan Ann ArborMI United States Department of Physics Harvard University CambridgeMA United States

Prior work has shown that text-conditioned diffusion models can learn to identify and manipulate primitive concepts underlying a compositional data-generating process, enabling generalization to entirely novel, out-of-distribution compositions. Beyond performance evaluations, these studies develop a rich empirical phenomenology of learning dynamics, showing that models generalize sequentially, respecting the compositional hierarchy of the data-generating process. Moreover, concept-centric structures within the data significantly influence a model’s speed of learning the ability to manipulate a concept. In this paper, we aim to better characterize these empirical results from a theoretical standpoint. Specifically, we propose an abstraction of prior work’s compositional generalization problem by introducing a structured identity mapping (SIM) task, where a model is trained to learn the identity mapping on a Gaussian mixture with structurally organized centroids. We mathematically analyze the learning dynamics of neural networks trained on this SIM task and show that, despite its simplicity, SIM’s learning dynamics capture and help explain key empirical observations on compositional generalization with diffusion models identified in prior work. Our theory also offers several new insights—e.g., we find a novel mechanism for non-monotonic learning dynamics of test loss in early phases of training. We validate our new predictions by training a text-conditioned diffusion model, bridging our simplified framework and complex generative models. Overall, this work establishes the SIM task as a meaningful theoretical abstraction of concept learning dynamics in modern generative models. © 2024, CC BY.

关键词： Mapping

Blockchain as a decentralized communication tool for sustainable development 2

学校读者我要写书评

暂无评论

Blockchain as a decentralized communication tool for sustain...

2nd International Conference on Smart Computing and Electronic Enterprise, ICSCEE 2021

作者： Baker El-Ebiary, Yousef A. Hatamleh, Amer Aseh, Khairi Altrad, Abdullah Amayreh, Khalid Thaher Hilles, Shadi Mohamed, Rajina R. Ravindran Pathmanathan, P. Aledinat, Lowai Saleh UniSZA Faculty of Informatics and Computing Malaysia Irbid National University Faculty of Financial and Business Science Information System Department Irbid Jordan International Institute of Applied Science Swiss School of Management Netherlands Al Madinah International University Faculty of Computer and Info. Tech Malaysia Irbid National University Faculty of Financial and Business Science Administration Department Irbid Jordan Istanbul Okan University Department of Software Engineering Turkey Universiti Tenaga Nasional College of Computing Dan Informatics Malaysia UniSZA Faculty of Informatics and Computing Malaysia

ISBN: (纸本)9781665432221

This research introduces the distributed ledger technology (blockchain) in terms of sustainable development in the future. This study is based on private, public and global communications in various industries. The growing demands for secure and decentralized transactions around the world demonstrate the need for Blockchain technology as a medium of operations and communications. The blockchain system is completely decentralized and allows users to exchange messages in an efficient and secure manner. The highlight of the paper is the anticipated future use of blockchain networks as a communication tool in every business and digital economy for a safe and sustainable development, while presenting its advantages and limitations that have been confirmed based on the experiences of various experts. Moreover, the article provides recommendations for the use of blockchain to greatly complicate its identification in future communications. The paper also concludes that the use of the blockchain system could be beneficial and enhance future communications and digital technologies. © 2021 IEEE.

关键词： Blockchain

Computing Safest $st$-Paths in Backbone Networks: Efficiently Solvable Cases and Fast Heuristics

学校读者我要写书评

暂无评论

Computing Safest $st$-Paths in Backbone Networks: Efficientl...

International Workshop on Reliable Networks Design and Modeling (RNDM)

作者： Balázs Vass Péter Revisnyei Alija Pašíć Department of Telecommunications and Artificial Intelligence (TMIT) Faculty of Electrical Engineering and Informatics (VIK) High Speed Networks Laboratory (HSNLab) Budapest University of Technology and Economics (BME) Budapest Hungary Faculty of Mathematics and Computer Science Babeș-Bolyai University Cluj Napoca Romania HUN-REN-BME Information Systems Research Group Budapest Hungary

ISBN: (数字)9798331518189

ISBN: (纸本)9798331518196

For proper evaluation and optimization of the expected availability of a backbone network service, two related very fundamental modeling and algorithmic questions are the following. 1) Realistically estimate the availability of a given $st$ -path for some source and target pair of communicating nodes $s$ and $t$ , and 2) Find a safest $st$ -path. For these, traditional approaches suppose network element failures are independent. In this paper, we show that, by not considering joint failure probabilities, the traditional approaches may misguess the $st$ -path availabilities and, consequently, the total connection availability, which can lead to more frequent Service Level Agreement (SLA) violations and a financial burden on the Communication Service Provider (CSP). Due to the inconsistent estimations, the supposedly safest paths yielded by these approaches may turn out to be suboptimal. On the positive side, we propose a fast algorithmic approach, that, under some assumptions, returns with a (truly) safest $st$ -path accompanied by its exact expected availability. When the aforementioned assumptions do not hold, in our simulations, the proposed algorithmic scheme proved to be a good heuristic, performing at least as well as the traditional method.

关键词： Computational modeling Heuristic algorithms Conferences Estimation Hazards Computational efficiency Optimization Service level agreements

RLCorrector: Reinforced Proofreading for Cell-level Microscopy Image Segmentation

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Nguyen, Khoa Tuan Jang, Ganghee Tuan, Tran Anh Jeong, Won-Ki Department of Computer Science & Engineering Korea University Korea Republic of Max Planck Institute for Informatics Saarland University Germany

Segmentation of nanoscale electron microscopy (EM) images is crucial but still challenging in connectomics research. One reason for this is that none of the existing segmentation methods are error-free, so they require proofreading, which is typically implemented as an interactive, semi-automatic process via manual intervention. Herein, we propose a fully automatic proofreading method based on reinforcement learning that mimics the human decision process of detection, classification, and correction of segmentation errors. We systematically design the proposed system by combining multiple reinforcement learning agents in a hierarchical manner, where each agent focuses only on a specific task while preserving dependency between agents. Furthermore, we demonstrate that the episodic task setting of reinforcement learning can efficiently manage a combination of merge and split errors concurrently presented in the input. We demonstrate the efficacy of the proposed system by comparing it with conventional proofreading methods over various testing cases. © 2021, CC BY-NC-ND.

关键词： Errors

Logic-consistency text generation from semantic parses

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Shu, Chang Zhang, Yusen Dong, Xiangyu Shi, Peng Yu, Tao Zhang, Rui School of Informatics University of Edinburgh Department of Computer Science and Engineering Penn State University School of Computer Science and Engineering Beihang University David R. Cheriton School of Computer Science University of Waterloo Department of Computer Science The University of Hong Kong

Text generation from semantic parses is to generate textual descriptions for formal representation inputs such as logic forms and SQL queries. This is challenging due to two reasons: (1) the complex and intensive inner logic with the data scarcity constraint, (2) the lack of automatic evaluation metrics for logic consistency. To address these two challenges, this paper first proposes SNOWBALL, a framework for logic consistent text generation from semantic parses that employs an iterative training procedure by recursively augmenting the training set with quality control. Second, we propose a novel automatic metric, BLEC, for evaluating the logical consistency between the semantic parses and generated texts. The experimental results on two benchmark datasets, Logic2Text and Spider, demonstrate the SNOWBALL framework enhances the logic consistency on both BLEC and human evaluation. Furthermore, our statistical analysis reveals that BLEC is more logically consistent with human evaluation than general-purpose automatic metrics including BLEU, ROUGE and, BLEURT. Our data and code are available at https://***/Ciaranshu/relogic. © 2021, CC BY.

关键词： Semantics

An Energy-Efficient Machine-Type Communication for Maritime Internet of Things 11

学校读者我要写书评

暂无评论

An Energy-Efficient Machine-Type Communication for Maritime ...

11th Annual IEEE Information Technology, Electronics and Mobile Communication Conference, IEMCON 2020

作者： Rahimi, Payam Chrysostomou, Chrysostomos Kyriakides, Ioannis Vassiliou, Vasos Frederick University Department of Electrical and Computer Engineering and Informatics Nicosia Cyprus University of Nicosia Research Foundation Nicosia Cyprus University of Cyprus and Rise Research Center Department of Computer Science Nicosia Cyprus

ISBN: (纸本)9781728184166

The Internet of Things (IoT) is an enabler technology for smart maritime networks. Connected IoT systems require reliable machine-Type communication (MTC). However, maritime MTC is facing several practical challenges including the wide-Area coverage, ubiquitous connectivity, cost-effectiveness, and reliability. In this paper, we first present a novel wireless communications assisted unmanned aerial vehicles (UAVs) system for maritime MTC. In our approach, UAVs are deployed to provide wide-Area coverage, while a network of connected buoys handles data transmission between the UAVs and the on-shore data fusion and control center (DFCC). We propose a handover decision method that eliminates unnecessary handover triggers;thus, reducing the overall energy consumption and ensuring seamless connectivity. We formulate the handover decision method as a constrained optimization problem of maximizing handover efficiency in terms of signal-To-noise ratio (SNR), available data rate, residual energy, and buffered data, by identifying the buoy offering optimal connectivity. The optimization problem is solved by a probabilistic based genetic algorithm (GA). We compare the proposed handover decision model with three benchmark scenarios to validate the performance gains achieved. © 2020 IEEE.

关键词： Genetic algorithms

The Smart Analysis of Performing Scalable Inference for Big Data Analytics

学校读者我要写书评

暂无评论

The Smart Analysis of Performing Scalable Inference for Big ...

International Conference on Power Energy, Environment and Intelligent Control (PEEIC)

作者： Vishakha D Bhandarkar Aruna Sri Rongali Mohamed Dawood Shamout Prof. Sachin Pund Gigih Forda Nama Shrishailappa Patil Applied Mathematics and Humanities Yeshwantrao Chavan College of Engineering India Graduate Doctoral Information Technology/Graduate (Information Technology Emphasis/Information Technology Ph.D.) University of The Cumberlands US Department of Management College of Business Administration University of Sharjah Sharjah United Arab Emirates Department of Industrial Engineering Shri Ramdeobaba College of Engineering and Management Nagpur Maharashtra India Department of Informatics Doctoral Program of Environtmental Science University of Lampung Bandar Lampung Computer engineering Vishwakarma Institute of Technology Pune

This paper examines the reproducibility of massive information analytics under particular factors. The paper proposes the “performing Scalable Inference” technique to cope with scalability troubles and to exploit current big statistics platforms for efficient computing and statistics garage of the statistics. In particular, the paper describes how to perform leak-free, parallelizable visible analytics over massive datasets using present extensive records analytics frameworks such as Apache Flink. This method presents an automated manner to execute analytics that preserves reproducibility and the ability to make adjustments without re-running the entire technique. The paper also demonstrates how these analytics may help several real-world use instances, explore affected person cohorts for studies, and develop stratified patient cohorts for hospital therapy. In the end, the paper observes how the proposed method may be exercised within the real world. Actively scalable inference for massive information analytics is pivotal in optimizing decision-making and allocation of assets. Typically, such inferences are made based on information accumulated from numerous sources, databases, unstructured data, and different digital sources. So one can ensure scalability, a complete cloud-primarily based platform has to be hired. This solution will permit the ***, deploying the essential records series and evaluation algorithms are prime here. It could permit the platform to recognize the styles inside the statistics and discover any ability correlations or traits. Additionally, predictive analytics and system mastering strategies may be incorporated to provide insights into the results of the information. In the long run, by leveraging those techniques, the platform can draw efficient inferences and appropriately compare situations in an agile and green way..

关键词：

Data Fusion in Neuromarketing: Multimodal Analysis of Biosignals, Lifecycle Stages, Current Advances, Datasets, Trends, and Challenges

学校读者我要写书评

暂无评论

SSRN

SSRN 2023年

作者： Pérez, Mario Quiles Beltrán, Enrique Tomás Martínez Bernal, Sergio López Prat, Eduardo Horna Del Campo, Luis Montesano Maimó, Lorenzo Fernández Celdrán, Alberto Huertas Department of Information and Communications Engineering University of Murcia Murcia30100 Spain Bit&Brain Technologies S.L Zaragoza50006 Spain Department of Computer Architecture and Technology University of Murcia Murcia30100 Spain Communication Systems Group CSG Department of Informatics IfI University of Zurich UZH ZürichCH—8050 Switzerland

The primary goal of any company is to increase its profits by improving both the quality of its products and how they are advertised. In this context, neuromarketing seeks to enhance the promotion of products and generate a greater acceptance on potential buyers. Traditionally, neuromarketing studies have relied on a single biosignal to obtain feedback from presented stimuli. However, thanks to new devices and technological advances studying this area of knowledge, recent trends indicate a shift towards the fusion of diverse biosignals. An example is the usage of electroencephalography for understanding the impact of an advertisement at the neural level and visual tracking to identify the stimuli that induce such impacts. This emerging pattern determines which biosignals to employ for achieving specific neuromarketing objectives. Furthermore, the fusion of data from multiple sources demands advanced processing methodologies. Despite these complexities, there is a lack of literature that adequately collates and organizes the various data sources and the applied processing techniques for the research objectives pursued. To address these challenges, the current paper conducts a comprehensive analysis of the objectives, biosignals, and data processing techniques employed in neuromarketing research. This study provides both the technical definition and a graphical distribution of the elements under revision. Additionally, it presents a categorization based on research objectives and provides an overview of the combinatory methodologies employed. After this, the paper examines primary public datasets designed for neuromarketing research together with others whose main purpose is not neuromarketing, but can be used for this matter. Ultimately, this work provides a historical perspective on the evolution of techniques across various phases over recent years and enumerates key lessons learned. © 2023, The Authors. All rights reserved.

关键词： Life cycle

Direction-assisted beam management in full duplex millimeter wave massive MIMO systems

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Islam, Md Atiqul Alexandropoulos, George C. Smida, Besma Department of Electrical and Computer Engineering University of Illinois Chicago United States Department of Informatics and Telecommunications National and Kapodistrian University of Athens Greece

Recent applications of the Full Duplex (FD) technology focus on enabling simultaneous control communication and data transmission to reduce the control information exchange overhead, which impacts end-to-end latency and spectral efficiency. In this paper, we present a simultaneous direction estimation and data transmission scheme for millimeter Wave (mmWave) massive Multiple-Input Multiple-Output (MIMO) systems, enabled by a recent FD MIMO technology with reduced hardware complexity Self-Interference (SI) cancellation. We apply the proposed framework in the mmWave analog beam management problem, considering a base station equipped with a large transmit antenna array realizing downlink analog beamforming and few digitally controlled receive antenna elements used for uplink Direction-of-Arrival (DoA) estimation. A joint optimization framework for designing the DoA-assisted analog beamformer and the analog as well as digital SI cancellation is presented with the objective to maximize the achievable downlink rate. Our simulation results showcase that the proposed scheme outperforms its conventional half-duplex counterpart, yielding reduced DoA estimation error and superior downlink data rate. © 2021, CC BY.

关键词： Millimeter waves