检索结果-内蒙古大学图书馆

arXiv 2024年

作者： Gould, Rhys Tanaka, Hidenori Department of Applied Mathematics and Theoretical Physics University of Cambridge United Kingdom CBS-NTT Physics of Intelligence Program Harvard University Physics & Informatics Laboratories NTT Research Inc United States

adaptive optimization algorithms, particularly Adam and its variant AdamW, are fundamental components of modern deep learning. However, their training dynamics lack comprehensive theoretical understanding, with limited insight into why common practices—such as specific hyperparameter choices and normalization layers—contribute to successful generalization. This work presents a continuous-time formulation of Adam and AdamW, facilitating a tractable analysis of training dynamics that can shed light on such practical questions. We theoretically derive a stable region for Adam’s hyperparameters (β, γ) that ensures bounded updates, empirically verifying these predictions by observing unstable exponential parameter growth outside of this stable region. Furthermore, we theoretically justify the success of normalization layers by uncovering an implicit meta-adaptive effect of scale-invariant architectural components. This insight leads to an explicit optimizer, 2-Adam, which we generalize to k-Adam—an optimizer that applies an adaptive normalization procedure k times, encompassing Adam (corresponding to k = 1) and Adam with a normalization layer (corresponding to k = 2). Overall, our continuous-time formulation of Adam facilitates a principled analysis, offering deeper understanding of optimal hyperparameter choices and architectural decisions in modern deep learning. © 2024, CC BY.

关键词： adaptive algorithms

来源：评论

学校读者我要写书评

暂无评论

Model Pruning Based on Large-Scale Multiobjective Evolution 3

Model Pruning Based on Large-Scale Multiobjective Evolution

引用

3rd International Conference on Artificial Intelligence, Human-Computer Interaction and Robotics, AIHCIR 2024

作者： Cao, Bin Du, Rui Hao, Yiming Liu, Weixuan School of Artificial Intelligence Hebei University of Technology Tianjin China National University of Singapore 21 Lower Kent Ridge Road Singapore

ISBN: (纸本)9798331534035

Neural network pruning can greatly reduce the computational demands of a model, allowing the model to be deployed in edge devices with limited arithmetic power. Previous pruning methods based on evolutionary algorithms have not only suffered accuracy loss but also neglected the internal correlation of the model to pursue a higher compression rate. To address these issues, this paper presents an adaptive large-scale multilevel competitive swarm evolutionary algorithm (ALMCSO). The algorithm comprehensively considers four optimization objectives—classification accuracy, the number of model parameters, FLOPs compression rate and structural similarity. It enhances global search capability through an adaptive hierarchical optimization strategy that dynamically balances convergence and diversity. A weighted optimization framework using a dynamic grouping strategy improves algorithm efficiency and reduces the search space dimension. Experiments on VGG and ResNet models using public image classification datasets show that ALMCSO enhances accuracy, achieves higher compression, and maintains model structural stability. © 2024 IEEE.

关键词： adaptive algorithms

来源：评论

学校读者我要写书评

暂无评论

Smart adaptive Beamforming 24

Smart Adaptive Beamforming

引用

Proceedings of the 7th International Conference on Networking, Intelligent Systems and Security

作者： Amellal Ra Rajaa Physic Department Laboratory Electronic Systems Information Processing Ibn Tofail Morocco

ISBN: (纸本)9798400709296

adaptive beamforming is a technique of signal processing that can ameliorate the capacity of the wireless communication and radar systems by reconfiguring the radiation pattern and maximizing the gain of the main lobe in a direction of arrival (DOA) of desired users to minimize side lobe and reducing signal to interference-plus-noise. adaptive algorithms Least Mean Square (LMS) operate the weight vectors of antenna array elements for beamforming by iterative process as well need to be continuously adapted to the ever-changing environment. Moreover, recursive least squares (RLS) give an advantage for fast convergence beamforming. The first phase for the smart beam former is used by direction of arrival (DoA) estimated using radial basis neural network (RBFNN).

关键词： adaptive algorithms

来源：评论

学校读者我要写书评

暂无评论

Low-Latency State Management for Real-Time Tasks in Edge Serverless

Low-Latency State Management for Real-Time Tasks in Edge Ser...

引用

International Symposium on Parallel and Distributed Processing with Applications, ISPA

作者： Yanyan Wen Guangping Xu Jianshe Wang Wei Hao School of Computer Science and Engineering Tianjin University of Technology China

ISBN: (数字)9798331509712

ISBN: (纸本)9798331509729

Stateful serverless systems commonly adopt an architectural paradigm characterized by compute and storage separation within cloud data centers. Nevertheless, guaranteeing prompt response for real-time tasks at the edge becomes challenging due to network overheads. This paper introduces a Low-Latency state management framework for real-time tasks in edge serverless systems called LoLa, which adaptively places states proximate to functions, thereby mitigating delays in accessing states within edge serverless systems. Our approach aims at mitigating network latency and optimizing resource utilization by co-locating functions and states, thereby enhancing the system’s overall efficiency. We introduce an adaptive strategy to coordinate the migration of states. It dynamically adjusts the positions of states based on historical data and real-time feedback. Additionally, we designed an in-memory state storage mechanism to facilitate low-latency access and implement a lightweight and fine-grained state management to ensure stored state consistency. Evaluation results showcase the efficacy of LoLa in reducing state read and write latency within edge serverless systems. Specifically, the average response latency is observed to decrease by 65.2% and 38.1% in the best and worst-case scenarios, respectively.

关键词： Technological innovation Distributed processing Data centers Heuristic algorithms Serverless computing adaptive algorithms Real-time systems Delays Low latency communication Optimization

来源：评论

学校读者我要写书评

暂无评论

Environmental adaptive Enhancement for the Bionic Polarized Compass Based on Multi-Scattering Light Model

SSRN

引用

SSRN 2024年

作者： Wang, Jue Hu, Pengwei Qian, Jianqiang Guo, Lei School of Physics Beihang University Beijing100191 China School of Automation Science and Electrical Engineering Beihang University Beijing100191 China National Key Laboratory on Aircraft Control Technology Beijing100191 China

The bionic polarized compass is an autonomous navigation technology with long-endurance capabilities that has recently garnered the attention of numerous researchers. However, the current algorithm of the polarized compass which is based on a single-scattering model demonstrates poor adaptability to environmental disturbances. Numerous academic studies have conclusively demonstrated that the multi-scattering model offers a more accurate approximation of the actual scenario. Inspired by insects, we found that the multi-scattering model has better environmental adaptability. However, the mathematical form of the multi-scattering model is often complex, making it difficult to directly obtain the solar vector from the polarized pattern. Therefore, we propose an inverse algorithm that combines the simulated annealing algorithm and a multi-scattering model, the equivalent incident light model(EIL model), to derive the solar vector from the polarized pattern with high environmental adaptability. Five experimental sets were conducted across diverse environments, revealing that the errors associated with the bionic polarized compass are consistently below 0.4°, representing a substantial improvement compared to existing compass technology. © 2024, The Authors. All rights reserved.

关键词： adaptive algorithms

来源：评论

学校读者我要写书评

暂无评论

Enhancing adaptive Feedback Cancellation Efficiency in Two-Microphone Hearing Aids

Enhancing Adaptive Feedback Cancellation Efficiency in Two-M...

引用

IEEE India Conference (INDICON)

作者： Parishmita Deka Bhargab Kalita Bijit Kumar Das Department of Electronics and Communication Engineering Indian Institute of Information Technology Guwahati Guwahati India

ISBN: (数字)9798350391282

ISBN: (纸本)9798350391299

The estimation of feedback paths and the mitigation of howling in hearing aids (HAs) depend heavily on adaptive feedback cancellation (AFC). Many techniques and approaches were studied over the decades to improve performance by reduce bias and increase performance. The most well-known and widely used approach is prediction error method-based AFC2 (PEM-AFC2). Improved estimation of the undesired feedback path is achieved by taking advantage of its inherent sparsity. This paper aims to improve the steady state misalignment value and tracking speed of PEM-AFC2 by incorporating two different sparsity exploiting techniques, like reweighted zero attraction (RZA) and hard-thresholding (HT), while maintaining signal quality. Performance metrics include speech quality assessment using Short-Time Objective Intelligibility (STOI) and Perceptual Evaluation of Speech Quality (PESQ), as well as misalignment (MIS) comparisons of the suggested algorithm with current algorithms.

关键词： adaptive systems Accuracy Prevention and mitigation Estimation adaptive algorithms Hearing aids Acoustic measurements Acoustics Steady-state Quality assessment

来源：评论

学校读者我要写书评

暂无评论

The Competition Complexity of Prophet Secretary

arXiv

引用

arXiv 2024年

作者： Ezra, Tomer Garbuz, Tamar Harvard University CambridgeMA United States Tel Aviv University Tel Aviv Israel

We study the classic single-choice prophet secretary problem through a resource augmentation lens. Our goal is to bound the (1 − Ε)-competition complexity for different classes of online algorithms. This metric asks for the smallest k such that the expected value of the online algorithm on k copies of the original instance, is at least a (1 − Ε)-approximation to the expected offline optimum on the original instance (without added copies). We consider four natural classes of online algorithms: single-threshold, time-based threshold, activation-based, and general algorithms. We show that for single-threshold algorithms the (1 − Ε)-competition complexity is Θ(ln(1/Ε)) (as in the i.i.d. case). Additionally, we demonstrate that time-based threshold and activation-based algorithms (which cover all previous approaches for obtaining competitive-ratios for the classic prophet secretary problem) yield a sub-optimal (1−Ε)-competition complexity of (Equation presented) which is strictly better than the class of single-threshold algorithms. Finally, we find that the (1 − Ε)competition complexity of general adaptive algorithms is (Equation presented) which is in sharp contrast to Θ(ln ln(1/Ε)) in the i.i.d. case. Copyright © 2024, The Authors. All rights reserved.

关键词： adaptive algorithms

来源：评论

学校读者我要写书评

暂无评论

adaptive Strategies and Handover Optimization in 5G Networks to Address High Mobility Challenges

Adaptive Strategies and Handover Optimization in 5G Networks...

引用

International Conference on Electrical Engineering and Informatics (ICELTICs)

作者： Murhaban Syahrial Teuku Yuliar Arif Fitri Arnia Departement Information Technology Engineering Doctoral Program School of Engineering Universitas Syiah Kuala Universitas Teuku Umar Banda Aceh Aceh Barat Indonesia Department of Electrical and Computer Engineering Universitas Syiah Kuala Banda Aceh Indonesia

ISBN: (数字)9798350366822

ISBN: (纸本)9798350366839

This study develops an adaptive handover strategy in 5G networks to tackle high mobility challenges and improve service continuity and quality. The research method uses a 5G network emulator with a topology of four base stations (gNBs) to test user equipment (UE) speeds of 20 km/h and 130 km/h. The adaptive algorithm developed utilizes data from the Received Signal Strength Indicator (RSSI) as well as the speed and direction of user mobility to make optimal handover decisions. The results show that at a speed of 130 km/h, the average handover latency reaches 12 ms, compared to 8 ms at 80 km/h. Additionally, the signal quality (RSSI) is lower at higher speeds, negatively affecting service quality. This adaptive algorithm successfully reduces handover latency and improves the stability and efficiency of the overall network. The testing also indicates that this adaptive algorithm can predict handover targets by considering UE speed and direction and uses hysteresis mechanisms to prevent overly frequent handovers. By employing technologies such as Multiple Input Multiple Output (MIMO) and beamforming, this strategy is expected to maintain high service quality even under high mobility conditions.

关键词： adaptive systems 5G mobile communication Network topology adaptive algorithms Handover Topology Received signal strength indicator Informatics Optimization Testing

来源：评论

学校读者我要写书评

暂无评论

adaptive Method for Finding the Decision-to-React Transition Point

Adaptive Method for Finding the Decision-to-React Transition...

引用

Neurotechnologies and Neurointerfaces (CNN), International Conference

作者： Nadezhda Titova Maria Pochueva Vladislav Aksiotis Alexey Ossadchi Alexey Tumyalis Centre for Bioelectric Interfaces HSE University Moscow Russia HSE University Moscow Russia Centre for Bioelectric Interfaces Federal center of brain research and neurotechnologies FMBA HSE University Moscow Russia

ISBN: (数字)9798350390926

ISBN: (纸本)9798350390933

Decision-making is a complex and extended in time process, hidden from direct measurement due to its interim location between the sensory perception of a stimulus and the preparing and execution of a motor reaction. Estimation of the duration of the decision-making process is relevant in fundamental research of cognitive and motor disorders, human-computer interfaces, cybersport and offers a promising avenue for advancing our understanding of decision-making mechanism in diverse populations. Determining the optimal transition point of decision-making processes to the execution of a motor program will allow solving these problems more effectively. In this study, we present an adaptive algorithm for estimating decision duration in a two-alternative forced-choice task. The subjects were presented with colored figures, during each trial the color of the figure has been changed, indicating the end of the preparation period (PP) and requirement for response execution. The task was performed in two conditions: constant (from 0 to 0.8 s) and adaptive PP durations. We found an initial decrease followed by stabilization of reaction time (RT) as the duration of constant PPs increases. The optimal duration of the preparation period was measured as the transition point of two regression lines fitted to the curve. Its location did not differ from mean adaptive PP and RT. Moreover, significant positive correlations were found between the results of the adaptive algorithm and the point of optimality. Thus, the adaptive algorithm localized the mean PP durations to the point of optimality. The proposed algorithm can be applied to determine the duration of decision-making and motor processes in the diagnosis of pathological conditions or used in brain-computer interfaces.

关键词： Human computer interaction Pathology Correlation Decision making adaptive algorithms Estimation Color Motors Time measurement Neurotechnology

来源：评论

学校读者我要写书评

暂无评论

An adaptive Three-Stage Algorithm For Solving Adjustable Min-Max-Regret Problems

arXiv

引用

arXiv 2024年

作者： Schneider, Kerstin Krieg, Helene Nowak, Dimitri Küfer, Karl-Heinz Fraunhofer Institute for Industrial Mathematics ITWM Kaiserslautern Germany

This work uniquely combines an affine linear decision rule known from adjustable robustness with min-max-regret robustness. By doing so, the advantages of both concepts can be obtained with an adjustable solution that is not over-conservative. This combination results in a bilevel optimization problem. For solving this problem, a three-stage algorithm which uses adaptive discretization of the uncertainty set via two criteria is presented and its convergence is proven. The algorithm is applicable for an example of optimizing a robust pump operation plan for a drinking water supply system facing uncertain demand. The algorithm shows a notable ability to scale, presenting an opportunity to solve larger instances that might challenge existing optimization *** Codes 90C17 © 2024, CC BY.

关键词： adaptive algorithms

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：