The computation of means, specifically the arithmetic mean, weighted mean, and harmonic mean, serves as a fundamental component in numerical computing, particularly in diverse signal processing applications. This pape...
详细信息
ISBN:
(数字)9798350361261
ISBN:
(纸本)9798350361261;9798350361278
The computation of means, specifically the arithmetic mean, weighted mean, and harmonic mean, serves as a fundamental component in numerical computing, particularly in diverse signal processing applications. This paper presents an adaptable, low-complexity, and reconfigurable scheme for the implementation of various mean calculations. The key concept involves utilizing in-memory computing (IMC) technique and executing computations through a set of memristive devices. Specifically, the required multiplications and additions leverage the inherent properties of memristor devices, adhering to Ohm's law and Kirchhoff's current law. This innovative approach not only enhances flexibility by leveraging the unique capabilities of IMC but also reduces computational complexity and latency compared to conventional implementation approaches.
distributed Order PID (DOPID) is a possible method, alternative to Fractional Order PID (FOPID), to extend the tuning options of the classical PID control by utilizing fractional calculus. With the classical FOPID sch...
详细信息
ISBN:
(纸本)9798331516246;9798331516239
distributed Order PID (DOPID) is a possible method, alternative to Fractional Order PID (FOPID), to extend the tuning options of the classical PID control by utilizing fractional calculus. With the classical FOPID scheme, the (PID mu)-D-lambda, two variable real orders, lambda and mu, are introduced and can be tuned to improve the system behaviour. On the contrary, with the DOPID approach fractional-order terms are added to the integer-order ones to extend the tuning possibilities. In particular, with the (PIIDD1/2)-D-1/2 scheme, the half-derivative and half-integral terms are added. Recently, a Bode plot- based method has been proposed for the (PIIDD1/2)-D-1/2 tuning, which derives the control parameters from those of a given PID without requiring optimization techniques. In the proposed work, this Bode-plot tuning method is extended to the PI lambda D-mu. Simulation results show that the proposed method, with the hypothesis of symmetry of the magnitude Bode plot, allows to achieve better performance in combination with the DOPID rather than with the FOPID.
As part of the Student Cluster Competition at the SC22 conference, this work aims to reproduce the performance evaluations of the Data Centric (DaCe) Python framework by leveraging Intel MKL and NVIDIA CUDA interface....
详细信息
As part of the Student Cluster Competition at the SC22 conference, this work aims to reproduce the performance evaluations of the Data Centric (DaCe) Python framework by leveraging Intel MKL and NVIDIA CUDA interface. The evaluations are conducted on a single CPU-based node, NVIDIA A100 GPUs, and an eight-node cloud supercomputer. Our experimental results successfully reproduce the performance evaluations on our cluster. Additionally, we provide insightful analysis and propose effective methods for achieving higher performance when utilizing DaCe as an acceleration library.
This paper designs and implements a deep learning based context-aware comic generation system, called StoryFrames, to automatically generate the comic strips with dialogues based on mobile cloud computing. The input t...
详细信息
ISBN:
(纸本)9798350363999;9798350364002
This paper designs and implements a deep learning based context-aware comic generation system, called StoryFrames, to automatically generate the comic strips with dialogues based on mobile cloud computing. The input text from a user to generate comic strips is analyzed to extract key contextual words for achieving artistic style consistency, which is a challenge faced by existing generative models. Through in-depth analysis of the input text, StoryFrames extracts visually and narratively rich scenes and ranks them based on their importance for arranging these scenes in the comic strips. In addition, deep sentiment analysis is explored to precisely reveal emotional fluctuations in the dialogues and select dialogue box styles that match the emotional states of the characters. To the best of our knowledge, this is the first system that explores deep learning, large language models, and mobile cloud computing to ensure coherence among elements within the comic strips and consistency in the overall story's logic, emotion, style, and contextual descriptions. In particular, an Android-based prototype is implemented to verify the feasibility and performance of StoryFrames.
In this paper, we consider the distributed time-varying optimization problem with coupled equality constraints over a connected undirected network. To address this issue, we design a novel distributed constraint optim...
详细信息
ISBN:
(纸本)9798350354416;9798350354409
In this paper, we consider the distributed time-varying optimization problem with coupled equality constraints over a connected undirected network. To address this issue, we design a novel distributed constraint optimization algorithm, and establish its ISS stability with external disturbances and tracking errors as the input and state, respectively. Moreover, the obtained result includes distributed constrained optimization with static objective functions as a special case. In comparison to existing relevant works, the proposed algorithm demonstrates exponential convergence for cases involving static objective functions. Finally, the theoretical results are validated via a numerical example.
The rapid rise in spatial data volumes from diverse sources necessitate efficient spatial data processing capability. Although most relational databases support spatial extensions of SQL query features, they offer lim...
详细信息
The rapid evolution of digital technologies and the pervasive nature of data connectivity have significantly expanded the scope of decentralized machine learning tasks. At the forefront of this shift is distributed ma...
详细信息
ISBN:
(纸本)9798331540913;9798331540906
The rapid evolution of digital technologies and the pervasive nature of data connectivity have significantly expanded the scope of decentralized machine learning tasks. At the forefront of this shift is distributed machine learning, which leverages distributed data while promoting privacy and efficiency. Built on the principles of cloud computing, distributed machine learning decomposes complex computational tasks into smaller components processed concurrently across interconnected nodes, optimizing resource utilization and scalability. The global cloud computing market, integral to the advancement of distributed machine learning, is projected to grow substantially, reaching USD 2,495.2 billion by 2032. Central to this study is the Cloud-Based Ratio Proportion Data Distribution Algorithm (CB-RPDDA), an innovative solution to traditional data distribution inefficiencies. CB-RPDDA reallocates data based on the processing speeds of individual machines, ensuring optimal resource utilization and effective workload distribution. This method introduces a new perspective on dataset division among worker nodes, enhancing load balancing and performance. By integrating CB-RPDDA with distributed machine learning frameworks, we improve the efficiency of decentralized learning processes, ensuring efficient data distribution across nodes while maintaining data security and privacy. Our findings demonstrate the potential of combining CB-RPDDA with distributed machine learning to offer scalable, efficient, and secure machine learning solutions, driving significant advancements in the field.
This paper investigates distributed time-varying optimization-based formation tracking problems for discrete-time heterogeneous multi-agent systems with unknown disturbances. Firstly, an optimization-based formation t...
详细信息
ISBN:
(纸本)9798350354416;9798350354409
This paper investigates distributed time-varying optimization-based formation tracking problems for discrete-time heterogeneous multi-agent systems with unknown disturbances. Firstly, an optimization-based formation tracking problem with privacy preservation is established, which formulates the relation between the formation tracking and the distributed optimization on the formation reference. Then, a distributed formation tracking controller composing of differential privacy mechanism and stochastic subgradient method is designed. Furthermore, the privacy, stability and optimality are proved by utilizing the discrete-time Lyapunov method. Finally, numerical simulations demonstrate the effectiveness of the proposed method.
The quantum cloud computing paradigm presents unique challenges in task placement due to the dynamic and heterogeneous nature of quantum computation resources. Traditional heuristic approaches fall short in adapting t...
详细信息
ISBN:
(纸本)9798350368543;9798350368536
The quantum cloud computing paradigm presents unique challenges in task placement due to the dynamic and heterogeneous nature of quantum computation resources. Traditional heuristic approaches fall short in adapting to the rapidly evolving landscape of quantum computing. This paper proposes DRLQ, a novel Deep Reinforcement Learning (DRL)-based technique for task placement in quantum cloud computing environments, addressing the optimization of task completion time and quantum task scheduling efficiency. It leverages the Deep Q Network (DQN) architecture, enhanced with the Rainbow DQN approach, to create a dynamic task placement strategy. This approach is one of the first in the field of quantum cloud resource management, enabling adaptive learning and decision-making for quantum cloud environments and effectively optimizing task placement based on changing conditions and resource availability. We conduct extensive experiments using the QSimPy simulation toolkit to evaluate the performance of our method, demonstrating substantial improvements in task execution efficiency and a reduction in the need to reschedule quantum tasks. Our results show that utilizing the DRLQ approach for task placement can significantly reduce total quantum task completion time by 37.81% to 72.93% and prevent task rescheduling attempts compared to other heuristic approaches.
distributed quantum computing (DQC) is a rapidly evolving field with its own unique challenges. Distributing a quantum algorithm involves several key steps and considerations. The steps involve decomposition at variou...
详细信息
ISBN:
(纸本)9798331541378
distributed quantum computing (DQC) is a rapidly evolving field with its own unique challenges. Distributing a quantum algorithm involves several key steps and considerations. The steps involve decomposition at various levels of abstraction, given the underlying quantum stack and quantum network capabilities. In our DQC design explorations, we focus on the distribution at the algorithm and circuit levels. Algorithmic distribution involves distributing tasks before compilation, allowing different quantum processing units (QPUs) to receive distinct parts of an algorithm. Circuit distribution involves executing a quantum algorithm in a distributed manner at the circuit execution level using circuit and adaptive quantum technologies. If entanglement across QPUs is supported, then quantum states can be shared between qubits on remote quantum processors. This requires a specialized architecture with data and communication qubits with non-local gates such as telegates and teledata gates. This paper presents our progress towards a framework for exploring quantum distribution at the algorithm and circuit levels. Our implementation and case studies demonstrate the feasibility of our approach and show effective pathways for distributed quantum algorithm experiments.
暂无评论