We present a 10-GHz hybrid actively and passively mode-locked fiber ring laser with a dispersion im-balanced nonlinear loop mirror (DI-NOLM), whose nonlinear switching characteristic can make the laser operate at addi...We present a 10-GHz hybrid actively and passively mode-locked fiber ring laser with a dispersion im-balanced nonlinear loop mirror (DI-NOLM), whose nonlinear switching characteristic can make the laser operate at additive pulse modelocking (APM) mode or additive pulse limiting (APL) mode. When the laser operated at APM, 5.45 ps of transform-limited pulse series were obtained. When the laser was biased at APL region, supermode noise was suppressed and the laser output was more stable.展开更多
基于强化学习(Reinforcement Learning,RL),在保证用户服务质量(Quality of Service,QoS)的前提下,研究了人机物混合接入的异构网络中通信-计算资源联合分配算法。建立了一种新型人机物混合接入的异构网络拓扑结构。在最小服务质量需求...基于强化学习(Reinforcement Learning,RL),在保证用户服务质量(Quality of Service,QoS)的前提下,研究了人机物混合接入的异构网络中通信-计算资源联合分配算法。建立了一种新型人机物混合接入的异构网络拓扑结构。在最小服务质量需求、无人机(Unmanned Aerial Vehicle,UAV)传输功率等限制条件下,将信道分配、功率分配和计算资源联合分配问题建模为最小化系统时延和能耗的多目标优化问题。基于强化学习理论和多智能体马尔可夫决策过程,提出一种分布式Q学习通信-计算资源联合分配(Distributed Q-learning Communication and Computing joint Resources Allocation,DQ-CCRA)算法。该算法与现有算法相比,不仅能够降低人类型设备对物类型设备的干扰,还能有效减小系统时延和能耗,将系统总开销降低7.4%。展开更多
文摘We present a 10-GHz hybrid actively and passively mode-locked fiber ring laser with a dispersion im-balanced nonlinear loop mirror (DI-NOLM), whose nonlinear switching characteristic can make the laser operate at additive pulse modelocking (APM) mode or additive pulse limiting (APL) mode. When the laser operated at APM, 5.45 ps of transform-limited pulse series were obtained. When the laser was biased at APL region, supermode noise was suppressed and the laser output was more stable.
文摘基于强化学习(Reinforcement Learning,RL),在保证用户服务质量(Quality of Service,QoS)的前提下,研究了人机物混合接入的异构网络中通信-计算资源联合分配算法。建立了一种新型人机物混合接入的异构网络拓扑结构。在最小服务质量需求、无人机(Unmanned Aerial Vehicle,UAV)传输功率等限制条件下,将信道分配、功率分配和计算资源联合分配问题建模为最小化系统时延和能耗的多目标优化问题。基于强化学习理论和多智能体马尔可夫决策过程,提出一种分布式Q学习通信-计算资源联合分配(Distributed Q-learning Communication and Computing joint Resources Allocation,DQ-CCRA)算法。该算法与现有算法相比,不仅能够降低人类型设备对物类型设备的干扰,还能有效减小系统时延和能耗,将系统总开销降低7.4%。