An Edge Server Placement Method Based on Reinforcement Learning

In mobile edge computing systems, the edge server placement problem is mainly tackled as a multi-objective optimization problem and solved with mixed integer programming, heuristic or meta-heuristic algorithms, etc. These methods, however, have profound defect implications such as poor scalability, local optimal solutions, and parameter tuning difficulties. To overcome these defects, we propose a novel edge server placement algorithm based on deep q-network and reinforcement learning, dubbed DQN-ESPA, which can achieve optimal placements without relying on previous placement experience. In DQN-ESPA, the edge server placement problem is modeled as a Markov decision process, which is formalized with the state space, action space and reward function, and it is subsequently solved using a reinforcement learning algorithm. Experimental results using real datasets from Shanghai Telecom show that DQN-ESPA outperforms state-of-the-art algorithms such as simulated annealing placement algorithm (SAPA), Top-K placement algorithm (TKPA), K-Means placement algorithm (KMPA), and random placement algorithm (RPA). In particular, with a comprehensive consideration of access delay and workload balance, DQN-ESPA achieves up to 13.40% and 15.54% better placement performance for 100 and 300 edge servers respectively.     

Learning Visible Thermal Person Re-Identification via Spatial Dependence and Dual-Constraint Loss

Visible thermal person re-identification (VT Re-ID) is the task of matching pedestrian images collected by thermal and visible light cameras. The two main challenges presented by VT Re-ID are the intra-class variation between pedestrian images and the cross-modality difference between visible and thermal images. Existing works have principally focused on local representation through cross-modality feature distribution, but ignore the internal connection of the local features of pedestrian body parts. Therefore, this paper proposes a dual-path attention network model to establish the spatial dependency relationship between the local features of the pedestrian feature map and to effectively enhance the feature extraction. Meanwhile, we propose cross-modality dual-constraint loss, which adds the center and boundary constraints for each class distribution in the embedding space to promote compactness within the class and enhance the separability between classes. Our experimental results show that our proposed approach has advantages over the state-of-the-art methods on the two public datasets SYSU-MM01 and RegDB. The result for the SYSU-MM01 is Rank-1/mAP 57.74%/54.35%, and the result for the RegDB is Rank-1/mAP 76.07%/69.43%.     

Modeling and Performance Optimization of an Irreversible Two-Stage Combined Thermal Brownian Heat Engine

Based on finite time thermodynamics, an irreversible combined thermal Brownian heat engine model is established in this paper. The model consists of two thermal Brownian heat engines which are operating in tandem with thermal contact with three heat reservoirs. The rates of heat transfer are finite between the heat engine and the reservoir. Considering the heat leakage and the losses caused by kinetic energy change of particles, the formulas of steady current, power output and efficiency are derived. The power output and efficiency of combined heat engine are smaller than that of single heat engine operating between reservoirs with same temperatures. When the potential filed is free from external load, the effects of asymmetry of the potential, barrier height and heat leakage on the performance of the combined heat engine are analyzed. When the potential field is free from external load, the effects of basic design parameters on the performance of the combined heat engine are analyzed. The optimal power and efficiency are obtained by optimizing the barrier heights of two heat engines. The optimal working regions are obtained. There is optimal temperature ratio which maximize the overall power output or efficiency. When the potential filed is subjected to external load, effect of external load is analyzed. The steady current decreases versus external load; the power output and efficiency are monotonically increasing versus external load.     

基于手性声子晶体的谷拓扑声输运

研究手性风车形散射体构建的谷拓扑声波导。具有右手与左手手性风车形散射体的声子晶体具有截然不同的声谷拓扑特性。当两种手性风车形散射体从-60°旋转到60°时,其所构建的声子晶体均出现2次偶然简并的狄拉克点与谷霍尔相变。基于两种具有相反谷霍尔相的手性声子晶体构建的谷拓扑声波导,在其重叠体带隙内存在一对局域在波导分界面处的谷态边缘态。实验研究表明,该边缘态可以很好的支持谷拓扑声输运,且对弯曲与无序两种缺陷具有一定的鲁棒性.     

基于光声光谱技术的NO,NO2气体分析仪研究

NO, NO₂是大气污染源中的常见气体, 对环境具有严重的危害性. 为检测污染源中这两种气体的浓度, 构建了成本较低的基于红外热辐射光源的光声光谱气体检测系统. 分析计算得到了NO, NO₂ 在2500–6667 nm波段吸收谱线. 通过建立光声传输线RLC振荡电路模型和仿真得到品质因数、声压大小与谐振腔长、内腔半径以及调制频率的关系, 据此设计了光声池几何结构. 实验表明该系统所测得的光声信号与气体浓度有很好的线性关系, 并且对NO, NO₂气体极限检测灵敏度分别达到4.01 和1.07 μL. 通过调节激光发射波长和选取滤波片, 该系统还可用于其他微量气体的浓度检测. 关键词： 大气污染 光声光谱 气体检测     

磁控溅射制备W掺杂VO2/FTO复合薄膜及其性能分析

为了获得相变温度低且热致变色性能优越的光学材料, 室温下在F:SnO₂ (FTO)导电玻璃基板表面沉积钨钒金属膜, 再经空气气氛下的热氧化处理, 制备了W掺杂VO₂/FTO复合薄膜, 利用X射线光电子能谱、X射线衍射和扫描电镜对薄膜的结构和表面形貌进行了分析. 结果表明: 高温热氧化处理过程中没有生成W, F, V混合氧化物, W以替换V原子的方式掺杂. 与采用相同工艺和条件制备的纯VO₂/FTO复合薄膜相比, W掺杂VO₂薄膜没有改变晶面取向, 仍具有(110)晶面择优取向, 相变温度下降到35 ℃左右, 热滞回线收窄到4 ℃, 高低温下的近红外光透过率变化量提高到28%. 薄膜的结晶程度明显提高, 表面变得平滑致密, 具有很好的一致性, 对光电薄膜器件的设计开发和工业化生产具有重要意义. 关键词： W掺杂 2')" href="#">VO₂ FTO导电玻璃 磁控溅射     

Evolution of geometrical structures,stabilities and electronic properties of neutral and anionic Li n Cu λ (n = 1–9, λ = 0, −1) clusters: compare with pure lithium clusters

The structural evolution, stabilities, and electronic properties of copper-doped lithium Li _n Cu^λ (n?=?1–9, λ?=?0, ?1) clusters have been systematically investigated using a density functional method at PW91PW91 level. Extensive searches for ground-state structures were carried out, and the results showed the copper tends to occupy the most highly coordinated position and form the largest probable number of bonds with lithium atoms. By calculating the binding energies per atom, fragmentation energies and the HOMO-LOMO gaps, we found LiCu, Li₇Cu, LiCu^?, Li₂Cu^? and Li₈Cu^? clusters have the stronger relative stability and enhanced chemical stability. The content and pattern of frontier MOs for the most stable doped isomers were analysed to investigate the bond nature of interaction among Li and Cu atoms. The results show some σ-type and π-type bonds are formed among them, and with small admixture of the Cu d characters. To achieve a deep insight into the electron localization and reliable electronic structure information, the natural population analysis and electron localization function were performed and discussed.     

Theory studies of the local lattice distortions and the EPR parameters for Cu2+, Mn2+ and Fe3+ centres in ZnWO4

The local lattice distortions and the electron paramagnetic resonance (EPR) parameters (g factors, hyperfine structure constants and zero-field splittings) for Cu²⁺, Mn²⁺ and Fe³⁺ in ZnWO₄ are theoretically studied based on the perturbation calculations for rhombically elongated octahedral 3d⁹ and 3d⁵ complexes. The impurity centres on Zn²⁺ sites undergo the local elongations of 0.01, 0.002 and 0.013 Å along the C₂ axis and the planar bond angle variations of 8.1°, 8.0° and 8.6° for Cu²⁺, Mn²⁺ and Fe³⁺, respectively, due to the Jahn–Teller effect and size and charge mismatch. In contrast to the host Zn²⁺ site with obvious axial elongation (～0.31 Å) and perpendicular (angular) rhombic distortion, all the impurity centres demonstrate more regular octahedral due to the above local lattice distortions. The copper centre exhibits significant Jahn–Teller reductions for the spin-orbit coupling and orbital angular momentum interactions, characterised by the Jahn–Teller reduction factor J (≈0.29 ? 1). The calculated EPR parameters agree well with the experimental results. The local structures of the impurity centres are analysed in view of the corresponding lattice distortions.     

Studies on controllability of directed networks with extremal optimization

Almost all natural, social and man-made-engineered systems can be represented by a complex network to describe their dynamic behaviors. To make a real-world complex network controllable with its desired topology, the study on network controllability has been one of the most critical and attractive subjects for both network and control communities. In this paper, based on a given directed–weighted network with both state and control nodes, a novel optimization tool with extremal dynamics to generate an optimal network topology with minimum control nodes and complete controllability under Kalman’s rank condition has been developed. The experimental results on a number of popular benchmark networks show the proposed tool is effective to identify the minimum control nodes which are sufficient to guide the whole network’s dynamics and provide the evolution of network topology during the optimization process. We also find the conclusion: “the sparse networks need more control nodes than the dense, and the homogeneous networks need fewer control nodes compared to the heterogeneous” (Liu et al., 2011  [18]), is also applicable to network complete controllability. These findings help us to understand the network dynamics and make a real-world network under the desired control. Moreover, compared with the relevant research results on structural controllability with minimum driver nodes, the proposed solution methodology may also be applied to other constrained network optimization problems beyond complete controllability with minimum control nodes.