Cloud-edge collaborative computing is a key technology towards delay-sensitive and computation-intensive applications in future cellular networks. this paper considers the combination of edge computing and cloud compu...
详细信息
ISBN:
(纸本)9798350374353;9798350374346
Cloud-edge collaborative computing is a key technology towards delay-sensitive and computation-intensive applications in future cellular networks. this paper considers the combination of edge computing and cloud computing, and studies the problem of task offloading in an "End-Edge-Cloud" collaborative architecture optimized for Quality of Experience (QoE) and system stability. To this end, this paper proposes a dual experience pool task offloading algorithm based on deep reinforcement learning and fuzzy logic (FDRL-DEP). Considering dynamic offloading requests and time-varying communication conditions, this paper models the problem as a Markov process and applys Dueling Deep Q-Network (Dueling DQN) to implement it. And this paper designs a two-stage fuzzy logic controller to improve the instability of Dueling DQN to the system when the parameters are not optimized. Via extensive simulation and theoretical analyses, this paper shows the effectiveness of the proposed FDRL-DEP framework on improving QoE and system stability.
this paper introduces the Multiple Greedy Quasi-Newton (MGSR1-SP) method, a novel approach to solving strongly-convex-strongly-concave (SCSC) saddle point problems. Our method enhances the approximation of the squared...
详细信息
ISBN:
(纸本)9798350377859;9798350377842
this paper introduces the Multiple Greedy Quasi-Newton (MGSR1-SP) method, a novel approach to solving strongly-convex-strongly-concave (SCSC) saddle point problems. Our method enhances the approximation of the squared indefinite Hessian matrix inherent in these problems, significantly improving both stability and efficiency through iterative greedy updates. We provide a thorough theoretical analysis of MGSR1-SP, demonstrating its linear-quadratic convergence rate. Numerical experiments conducted on AUC maximization and adversarial debiasing problems, compared with state-of-the-art algorithms, underscore our method's enhanced convergence rate. these results affirm the potential of MGSR1-SP to improve performance across a broad spectrum of machine learningapplications where efficient and accurate Hessian approximations are crucial.
Voltage control in low-observable distribution networks faces significant uncertainty challenges. In this paper, an uncertainty-aware voltage control method based on the fusion of Bayesian deep learning and probabilis...
详细信息
the rapid growth of embedded systems has led to the integration of artificial intelligence capabilities at the edge of computer devices. the Beagle-Bone AI and AI64 platform is a potent development board intended to m...
详细信息
the proceedings contain 65 papers. the topics discussed include: U-shape spatial-temporal prediction network based on 3D convolution and BDLSTM;citywide PM2.5 concentration prediction using deep learning model;Swin-CA...
ISBN:
(纸本)9798350374346
the proceedings contain 65 papers. the topics discussed include: U-shape spatial-temporal prediction network based on 3D convolution and BDLSTM;citywide PM2.5 concentration prediction using deep learning model;Swin-CANet: a novel integration of Swin transformer with channel attention for enhanced motor imagery classification;a combined LSTM-CNN model for abnormal electricity usage detection;empowerment of large language models in psychological counseling through prompt engineering;diffusion vision transformer of temporomandibular joint classification;a chatbot for learning and life guide in the university based on LangChain+ChatGLM;repairing neural networks for image classification problems using spectrum-based fault localization;and practical workflows to engineer scalable presentation platforms for modern web applications.
the proceedings contain 40 papers. the topics discussed include: application of genetics algorithms in cryptography;adventures in deep learning-assisted multimodality medical imaging wonderland;topological analysis of...
ISBN:
(纸本)9798350367591
the proceedings contain 40 papers. the topics discussed include: application of genetics algorithms in cryptography;adventures in deep learning-assisted multimodality medical imaging wonderland;topological analysis of deep network training;enhancing industry 4.0 maturity: integrating lean practices as a key feature in artificial intelligence-driven decision support model for companies;fuzzy linguistic signatures and their applications;artificial intelligence tools for public relations practitioners: an overview;comparative analysis of machine learningalgorithms in traffic mainstream control on freeway networks;fortifying decentralized governance: introducing decentralized autonomous verification (DAVe) for decentralized autonomous organization (DAOs) security;and nature inspired optimizationalgorithms in fractional order controller design.
the dynamic job shop scheduling problem is crucial in modern manufacturing industries, especially in parallel batch processing environments where achieving multi-objective optimization still faces great challenges. In...
详细信息
Using mobile edge computing (MEC) servers to cache and update video resources can not only save network bandwidth and computing resources, but also adjust the bit rate of video playback according to real-time network ...
详细信息
the intelligent embedded system is a computing system combining artificial intelligence science and embedded technology, the system combines general-purpose processor and FPGA to achieve stronger processing capability...
详细信息
ISBN:
(纸本)9789819603534;9789819603541
the intelligent embedded system is a computing system combining artificial intelligence science and embedded technology, the system combines general-purpose processor and FPGA to achieve stronger processing capability, but it also brings system hardware and software design challenges. this paper proposes an intelligent embedded system task scheduling algorithm based on heterogeneous multi-core, which reasonably allocates hardware resources on the FPGA to optimize energy consumption under the premise of meeting system reliability requirements. the algorithm adopts a critical path-based energy consumption pre-allocation strategy, and the task scheduling makes the system schedule length shortest based on system reliability. Experiments show that the method of this paper outperforms other algorithms by an average of 5.34% in terms of energy consumption, and by an average of 6.83% in terms of scheduling length, which reflects the reasonableness and advancement of the method of this paper.
In machine learningapplications, it is well known that carefully designed learning rate (step size) schedules can significantly improve the convergence of commonly used first-order optimizationalgorithms. therefore ...
详细信息
In machine learningapplications, it is well known that carefully designed learning rate (step size) schedules can significantly improve the convergence of commonly used first-order optimizationalgorithms. therefore how to set step size adaptively becomes an important research question. A popular and effective method is the Polyak step size, which sets step size adaptively for gradient descent or stochastic gradient descent without the need to estimate the smoothness parameter of the objective function. However, there has not been a principled way to generalize the Polyak step size for algorithms with momentum accelerations. this paper presents a general framework to set the learning rate adaptively for first-order optimization methods with momentum, motivated by the derivation of Polyak step size. It is shown that the resulting techniques are much less sensitive to the choice of momentum parameter and may avoid the oscillation of the heavy-ball method on ill-conditioned problems. these adaptive step sizes are further extended to the stochastic settings, which are attractive choices for stochastic gradient descent with momentum. Our methods are demonstrated to be more effective for stochastic gradient methods than prior adaptive step size algorithms in large-scale machine learning tasks.
暂无评论