Quantum machine learning for Lyapunov-stabilized computation offloading in next-generation MEC networks

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Янв. 2, 2025

Quantum computing and machine learning convergence enable powerful new approaches for optimizing mobile edge (MEC) networks. This paper uses Lyapunov optimization theory to propose a novel quantum framework stabilizing computation offloading in next-generation MEC systems. Our approach leverages hybrid quantum-classical neural networks learn optimal policies that maximize network performance while ensuring the stability of data queues, even under dynamic unpredictable conditions. Rigorous mathematical analysis proves our controller achieves close-to-optimal bounding queue backlogs. Extensive simulations demonstrate proposed significantly outperforms conventional approaches, improving throughput by up 30% reducing power consumption over 20%. These results highlight immense potential revolutionize support emerging applications at intelligent edge.

Язык: Английский

---

Development of Organic Semiconductor Materials for Organic Solar Cells via the Integration of Computational Quantum Chemistry and AI-Powered Machine Learning

ACS Applied Energy Materials, Год журнала: 2025, Номер unknown

Опубликована: Янв. 13, 2025

Язык: Английский

---

Deep reinforcement learning-based scheduling in distributed systems: a critical review

Knowledge and Information Systems, Год журнала: 2024, Номер 66(10), С. 5709 - 5782

Опубликована: Июнь 26, 2024

Язык: Английский

---

A transformer-based deep reinforcement learning approach for dynamic parallel machine scheduling problem with family setups

Journal of Intelligent Manufacturing, Год журнала: 2024, Номер unknown

Опубликована: Авг. 8, 2024

Abstract The parallel machine scheduling problem (PMSP) involves the optimized assignment of a set jobs to collection machines, which is proper formulation for modern manufacturing environment. Deep reinforcement learning (DRL) has been widely employed solve PMSP. However, majority existing DRL-based frameworks still suffer from generalizability and scalability. More specifically, state action design heavily rely on human efforts. To bridge these gaps, we propose practical learning-based framework tackle PMSP with new job arrivals family setup constraints. We variable-length matrix containing full information. This enables DRL agent autonomously extract features raw data make decisions global perspective. efficiently process this novel matrix, elaborately modify Transformer model represent agent. By integrating modified agent, representation can be effectively leveraged. innovative offers high-quality robust solution that significantly reduces reliance manual effort traditionally required in tasks. In numerical experiment, stability proposed during training first demonstrated. Then compare trained 192 instances several approaches, namely approach, metaheuristic algorithm, dispatching rule. extensive experimental results demonstrate scalability our approach its effectiveness across variety scenarios. Conclusively, thus problems high efficiency flexibility, paving way application solving complex dynamic problems.

Язык: Английский

---

An Energy-Efficient Dynamic Offloading Algorithm for Edge Computing Based on Deep Reinforcement Learning

IEEE Access, Год журнала: 2024, Номер 12, С. 127489 - 127506

Опубликована: Янв. 1, 2024

Язык: Английский

---

A group scheduling algorithm for massive heterogeneous data in the “dual carbon” digital intelligence monitoring center considering time-varying characteristics and priorities

Sustainable Energy Research, Год журнала: 2025, Номер 12(1)

Опубликована: Янв. 24, 2025

Язык: Английский

---

Scaling Up Optuna: P2P Distributed Hyperparameters Optimization

Concurrency and Computation Practice and Experience, Год журнала: 2025, Номер 37(4-5)

Опубликована: Фев. 28, 2025

ABSTRACT In machine learning (ML), hyperparameter optimization (HPO) is the process of choosing a tuple values that ensures an efficient deployment and training AI model. practice, HPO not only applies to ML tuning but can also be used tune complex numerical simulations. this context, model given object created in realistic This defined by set describing properties such as geometry or other unknown parameters related physical quantities. While for usually requires finding few parameters, involve more than hundred parameters. As consequence, large number tuples have explored evaluated before relevant solution, offering new challenges high‐performance computing efficiently driving optimization. work we rely on Optuna framework, primarily designed tasks including state‐of‐the‐art sampling pruning algorithms. We report its use optimize onto 1024‐core machine. suggest 1.5M evaluate 5M simulations using different Optuna‐distributed layouts build several tradeoffs between performance energy consumption metrics. order further scale up resources, introduce OptunaP2P, extension based peer‐to‐peer paradigm. allows remove any bottleneck management shared knowledge processes. With were able compute 3 times faster compared regular implementation obtain close‐to‐similar results terms quality reduced time‐frame.

Язык: Английский

---

Reinforcement learning for data center energy efficiency optimization: A systematic literature review and research roadmap

Applied Energy, Год журнала: 2025, Номер 389, С. 125734 - 125734

Опубликована: Март 26, 2025

Язык: Английский

---

Towards Efficient Job Scheduling for Cumulative Data Processing in Multi-Cloud Environments

Electronics, Год журнала: 2025, Номер 14(7), С. 1332 - 1332

Опубликована: Март 27, 2025

The rapid expansion of multi-cloud environments enables the fulfillment dynamic and diverse resource requirements cloud applications. Cumulative data processing (CDP) applications, which handle incrementally generated in stages like preprocessing aggregate analysis, particularly benefit from these environments. However, existing scheduling solutions struggle to accumulation processed long-term operation dependencies CDP Aiming at this issue, we propose a novel job execution model, CDP-EM, tailored strategy, CDP-JS, optimize applications CDP-EM model generation dependency-aware for while CDP-JS strategy formulates problem as Markov Decision Process (MDP), utilizing deep reinforcement learning with Proximal Policy Optimization (PPO) decisions. simulation results show that integrating reduces SLA violation rate cost by an average 34.8% 23.4%, respectively. Real-world evaluations reductions 27.2% 31.3%,

Язык: Английский

---

A multi-objective approach to load balancing in cloud environments integrating ACO and WWO techniques

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Апрель 8, 2025

Effective load balancing and resource allocation are essential in dynamic cloud computing environments, where the demand for rapidity continuous service is perpetually increasing. This paper introduces an innovative hybrid optimisation method that combines water wave optimization (WWO) ant colony (ACO) to tackle these challenges effectively. ACO acknowledged its proficiency conducting local searches effectively, facilitating swift discovery of high-quality solutions. In contrast, WWO specialises global exploration, guaranteeing extensive coverage solution space. Collectively, methods harness their distinct advantages enhance various objectives: decreasing response times, maximising efficiency, lowering operational expenses. We assessed efficacy our methodology by simulations using a cloud-sim simulator variety workload trace files. comparison well-established algorithms, such as WWO, genetic algorithm (GA), spider monkey (SMO), ACO. Key performance indicators, task scheduling duration, execution costs, energy consumption, utilisation, were meticulously assessed. The findings demonstrate WWO-ACO approach enhances efficiency 11%, decreases expenses 8%, lowers usage 12% relative conventional methods. addition, consistently achieved impressive equilibrium allocation, with balance values ranging from 0.87 0.95. results emphasise algorithm's substantial impact on improving system customer satisfaction, thereby demonstrating significant improvement techniques.

Язык: Английский

---