A5100621613- Advanced MIMO Systems Optimization
- Advanced Numerical Analysis Techniques
- Advanced Fiber Optic Sensors
- Cooperative Communication and Network Coding
- Wireless Communication Security Techniques
- Photonic and Optical Devices
- Mathematics and Applications
- 3D Shape Modeling and Analysis
- Injection Molding Process and Properties
- Potato Plant Research
- Advanced Data and IoT Technologies
- Human Mobility and Location-Based Analysis
- Plasmonic and Surface Plasmon Research
- Melamine detection and toxicity
- Sulfur Compounds in Biology
- Advanced Surface Polishing Techniques
- Microwave Engineering and Waveguides
- Aquatic Ecosystems and Phytoplankton Dynamics
- Fish Biology and Ecology Studies
- Indoor and Outdoor Localization Technologies
- Magneto-Optical Properties and Applications
- Additive Manufacturing and 3D Printing Technologies
- Laser Material Processing Techniques
- Computational Geometry and Mesh Generation
- Advanced Image and Video Retrieval Techniques
Anhui University
2022-2023
State Key Laboratory of Food Science and Technology
2023
Jiangnan University
2023
Stanford University
2022
Nanjing University of Information Science and Technology
2022
Chinese Academy of Sciences
2021
Shenyang Institute of Automation
2021
Nanyang Technological University
2017
Nanjing University of Science and Technology
2015
当已知未来的移动轨迹等信息时,面向视频点播业务的预测资源分配可以在保证移动用户体验的前提下降低基站能耗或提高网络吞吐量。 传统的预测资源分配方法采用先预测用户轨迹等信息再优化功率等资源分配的方法,在预测窗较长时预测误差大,导致预测所带来的增益降低。为了解决这个问题,近期已有文献把预测资源分配建模为马尔科夫决策过程、采用深度强化学习进行在线决策。 然而,对于这类适于采用强化学习的马尔科夫决策过程,现有文献往往以试错的方式对状态进行设计。此外,对于有约束的优化问题,现有利用强化学习解决无线问题的方法大多通过在奖励函数上加入包含需要手动调节超参数的惩罚项满足约束。本文以移动用户视频播放不卡顿约束下使基站发射能耗最小的目标为例,提出了端到端在线求解预测资源分配的无监督深度学习方法对信息预测和资源分配进行联合优化,能以系统化的方式设计状态且满足约束,建立了这种方法与深度强化学习的联系。仿真结果表明,所提出的在线无监督深度学习与深度强化学习所达到的发射能耗相近,但能够简化状态的设计,验证了理论分析结果。
Hybrid precoding can support high data rate with low cost for millimeter wave (mmWave) multi-antenna systems. To achieve near-optimal performance computing latency and enable end-to-end learning, deep learning has been introduced optimizing hybrid precoding. Most research efforts focus on under static channels. In mobile communications, however, the channel aging effect incurs severe degradation of this paper, we resort to a proactive optimization method learn analog digital precoders...
Alumina microlattices with stretch- and bending-dominated structures were manufactured to establish the intercorrelation between quasi-static microlattice mechanical properties strut aspect ratio using stereolithographic additive manufacturing method. In present work, two kinds of novel prediction models calibrated validated through experiment simulation data, aiming realize topological optimization effective design weight, size, spatial geometrical according given load bearing stiffness...
Quenching protection is one of the key challenges for operation high-temperature superconductor (HTS) magnets, and optical frequency domain reflection (OFDR) technology has great application potential in quenching detection HTS due to its high spatial resolution, fast response, anti-electromagnetic interference, other characteristics. In this article, we monitored temperature change process simulated device by combining fiber weak point phase demodulation OFDR technology, finally achieved...
Improving learning efficiency is paramount for resource allocation with deep neural networks (DNNs) in wireless communications over highly dynamic environments. Incorporating domain knowledge into a promising way of dealing this issue, which an emerging topic the community. In article, we first briefly summarize two classes approaches to using knowledge: introducing mathematical models or prior learning. Then, consider kind symmetric prior, permutation equivariance, widely exists tasks. To...
Abstract A novel design for a highly-sensitive plasmonic sensor, consisting of metal-insulator-metal (MIM) waveguide with coupled split-square-ring resonator, is presented and studied by two-dimensional (2D) finite-difference time-domain (FDTD) method in this paper. The simulation results show that the proposed sensor possesses refractive index (RI) sensitivity up to 2040 nm/RIU optimized structural parameters. In addition, temperature −1.2 nm/◦C determined. Numerical calculations indicate...