Mitigating Straggler Effect in Federated Learning Based on Reconfigurable Intelligent Surface over Internet of Vehicles

To protect vehicular privacy and speed up the execution of tasks,federated learning is introduced in the Internet of Vehicles(IoV)where users execute model training locally and upload local models to the base station without massive raw data exchange.However,heterogeneous computing and communication resources of vehicles cause straggler effect which weakens the reliability of federated learning.Dropping out vehicles with limited resources confines the training data.As a result,the accuracy and applicability of federated learning models will be reduced.To mitigate the straggler effect and improve performance of federated learning,we propose a reconfigurable intelligent surface(RIS)-assisted federated learning framework to enhance the communication reliability for parameter transmission in the IoV.Furthermore,we optimize the phase shift of RIS to achieve a more reliable communication environment.In addition,we define vehicular competence to measure both vehicular trustworthiness and resources.Based on the vehicular competence,the straggler effect is mitigated where training tasks of computing stragglers are offloaded to surrounding vehicles with high competence.The experiment results verify that our proposed framework can improve the reliability of federated learning in terms of computing and communication in the IoV.

S1280 and S1284: Two Oscillating Blue Stragglers in the Open Cluster M67

We present results of a time-series CCD photometry of two blue stragglers in the open cluster M67 that are also oscillating variables, S1280 and S1284. The observations obtained on 11 nights confirmed the δ Scuti-like variability of the two stars. Four and five main pulsating frequencies are detected for S1280 and S1284, respectively, through a power spectral analysis. A preliminary mode identification indicates that the two stars are both in radial oscillation. Based on the nature of oscillation, the physical parameters of the two stars are determined, and their evolutionary status discussed.

Modeling blue stragglers in young clusters

A grid of binary evolution models are calculated for the study of a blue straggler （BS） population in intermediate age （log Age = 7.85 - 8.95） star clusters. The BS formation via mass transfer and merging is studied systematically using our models. Both Case A and B close binary evolutionary tracks are calculated for a large range of parameters. The results show that BSs formed via Case B are generally bluer and even more luminous than those produced by Case A. Furthermore, the larger range in orbital separations of Case B models provides a probability of producing more BSs than in Case A. Based on the grid of models, several Monte-Carlo simulations of BS populations in the clusters in the age range are carded out. The results show that BSs formed via different channels populate different areas in the color magnitude diagram （CMD）. The locations of BSs in CMD for a number of clusters are compared to our simulations as well. In order to investigate the influence of mass transfer efficiency in the models and simulations, a set of models is also calculated by implementing a constant mass transfer efficiency,β=0.5, during Roche lobe overflow （Case A binary evolution excluded）. The result shows BSs can be formed via mass transfer at any given age in both cases. However, the distributions of the BS populations on CMD are different.

The fractions of post-binary-interaction stars and evolved blue straggler stars on the red giant branch of globular clusters

The red giant branch(RGB)of globular clusters(GCs)is home to some exotic stars,which may provide clues on the formation of multiple stellar populations in GCs.It is well known that binary interactions are responsible for many exotic stars.Thus,it is important to understand what fraction of stars on the RGB of GCs is the result of binary interactions.In this paper,we performed a binary population synthesis study to track the number of post-binary-interaction(post-BI)stars that appear on the RGB,with particular emphasis on the evolved blue straggler stars(E-BSSs).Assuming an initial binary fraction of nearly 50%,we find that about half of the objects on the RGB(called giants)underwent the binary interactions,and that E-BSSs account for around 10%of the giants in our standard simulation.We also compare the properties of post-BI giants that evolved from different channels.We find that the initial orbital period and mass ratio distributions significantly affect the fraction of post-BI giants.Our results imply that the non-standard stars from binary interactions provide a non-negligible contribution to the RGB stars in GCs,which should be considered in future investigations of the origin of multiple stellar populations.

The Blue Stragglers and the Integrated Spectral Energy Distribution of Single Stellar Populations

Blue stragglers are a common observational fact for the Galactic clusters. Single Stellar Populations (SSPs) are basic to the studies of galaxy structre and evolution. SSPs are mainly based either on the observation of the integrated properties of star clusters, or on the theoretical understandings of single star evolution. Both of the two ways of making SSPs suffer from either observational uncertainties concerning field contaminations or lack of good models for close binary systems. Based on the photometry of the classical open cluster M67 and the thorough membership survey, we made a color-magnitude diagram (CMD) of high membership stars for the cluster. We will show that by including the contributions of the bright blue stragglers that is common to open clusters, the integrated properties of the clusters are quite different from tranditional SSP models. We further conclude that these blue light contributors are very important to SSP models, and may cast new lights on its applications in the studies of galaxies.

Partial Clones for Stragglers in MapReduce

Stragglers can temporize jobs and reduce cluster efficiency seriously.Many researches have been contributed to the solution, such as Blacklist[8], speculative execution[1, 6], Dolly[8]. In this paper, we put forward a new approach for mitigating stragglers in MapReduce, name Hummer. It starts task clones only for high-risk delaying tasks. Related experiments have been carried and results show that it can decrease the job delaying risk with fewer resources consumption. For small jobs, Hummer also improves job completion time by 48% and 10% compared to LATE and Dolly.

Elastic Optimization for Stragglers in Edge Federated Learning

To fully exploit enormous data generated by intelligent devices in edge computing,edge federated learning(EFL)is envisioned as a promising solution.The distributed collaborative training in EFL deals with delay and privacy issues compared to traditional centralized model training.However,the existence of straggling devices,responding slow to servers,degrades model performance.We consider data heterogeneity from two aspects:high dimensional data generated at edge devices where the number of features is greater than that of observations and the heterogeneity caused by partial device participation.With large number of features,computation overhead on the devices increases,causing edge devices to become stragglers.And incorporation of partial training results causes gradients to be diverged which further exaggerates when more training is performed to reach local optima.In this paper,we introduce elastic optimization methods for stragglers due to data heterogeneity in edge federated learning.Specifically,we define the problem of stragglers in EFL.Then,we formulate an optimization problem to be solved at edge devices.We customize a benchmark algorithm,FedAvg,to obtain a new elastic optimization algorithm(FedEN)which is applied in local training of edge devices.FedEN mitigates stragglers by having a balance between lasso and ridge penalization thereby generating sparse model updates and enforcing parameters as close as to local optima.We have evaluated the proposed model on MNIST and CIFAR-10 datasets.Simulated experiments demonstrate that our approach improves run time training performance by achieving average accuracy with less communication rounds.The results confirm the improved performance of our approach over benchmark algorithms.

Photometric analysis of a blue straggler eclipsing binary in the old open cluster NGC 2141

To understand the formation mechanism of blue stragglers in open clusters, we conducted a photometric analysis for a blue straggler eclipsing binary in the old open cluster NGC 2141. By using the least square fitting of times of minima, we derived the ephemeris and redetermined the orbital period of this system to be 0.6227 d. We have also done a photometric analysis for light curve with the Wilson-Devinney Code （2003）. The photometric solutions reveal a semi-detached binary with the second- ary component filling its Roche lobe. The photometric mass ratio MJMI （secondary/primary） was determined to be 0.190_＋0.002 and the ratio of V-band luminosity L2/LI（V） is about 0.033_＋0.001, which indicates the secondary component is overluminous. Thus, this blue straggler likely results from a substantial mass exchange between these two components.

面向异构分布式机器学习的动态自适应并行加速方法

分布式机器学习因其优越的并行能力成为人工智能领域复杂模型训练的常用技术。然而,GPU升级换代非常快,异构集群环境下的分布式机器学习成为数据中心、研究机构面临的新常态。异构节点之间训练速度的差异使得现有并行方法难以平衡同步等待和陈旧梯度的影响,从而显著降低模型整体训练效率。针对该问题,提出了一种基于节点状态的动态自适应并行方法(dynamic adaptive synchronous parallel,DASP),利用参数服务器动态管理节点训练时的状态信息并对节点的并行状态进行划分,通过节点状态信息自适应调整每个节点的并行状态,以减少快速节点对全局模型参数的同步等待时间与陈旧梯度的产生,从而加快收敛效率。在公开数据集上的实验结果表明,DASP比主流方法收敛时间减少了16.9%~82.1%,并且训练过程更加稳定。

基于任务进度感知的异构Hadoop云平台任务调度方案

针对异构Hadoop云计算平台中任务调度问题进行了研究,并分析了现有推测执行调度和LATE调度方案的缺陷,提出一种基于任务进度感知的自适应任务调度方案。首先,根据当前计算节点上的任务进度情况,估计任务近似完成时间(ATE),以此确定掉队者(straggler)任务;然后,以平均节点任务进度的25%为阈值,将节点分为慢节点和快节点。当straggler后备任务达到一定阈值时,将其优先分配到快节点中执行。实验结果表明,提出的方案能够为异构Hadoop平台合理分配任务,有效降低了任务完成时间和响应延迟,同时提高了平台吞吐量。

小学数学学困生的学习情感心理特征及影响因素研究

小学数学学困生指的是智力正常,但学习效果低下,学业成绩明显落后于同年龄或同年级的水平的学生.小学数学学困生的学习情感心理特征与其学业成绩显著地负相关,小学数学学困生的学习情感心理特征性别差异不显著,年级与城乡差异均显著.

编码技术改进大规模分布式机器学习性能综述

由于分布式计算系统能为大数据分析提供大规模的计算能力,近年来受到了人们的广泛关注.在分布式计算系统中,存在某些计算节点由于各种因素的影响,计算速度会以某种随机的方式变慢,从而使运行在集群上的机器学习算法执行时间增加,这种节点叫作掉队节点(straggler).介绍了基于编码技术解决这些问题和改进大规模机器学习集群性能的研究进展.首先介绍编码技术和大规模机器学习集群的相关背景;其次将相关研究按照应用场景分成了应用于矩阵乘法、梯度计算、数据洗牌和一些其他应用,并分别进行了介绍分析;最后总结讨论了相关编码技术存在的困难并对未来的研究趋势进行了展望.

Hadoop云平台下基于LATE改进的推测执行算法

为解决Hadoop云平台下推测执行算法的不足,基于LATE算法提出一种新的任务进度计算方法,并且采用更细粒度的方式选择执行备份任务的节点。实验结果表明,该算法能够更精准的选定掉队者任务并选择合适的执行节点,缩短作业的执行时间,提高云平台的效率。

Two particular EA-type binaries in the globular cluster ω Centauri

We analyzed the B and V light curves of two EA-type binaries V211 and NV358 for the first time using the Wilson-Devinney code. Our analysis shows that V211 is a typical Algol-type binary and NV358 is a well detached binary system. As the two binaries are definite members of ω Centauri due to their proper motion, we estimated their physical parameters, obtaining M1 = 1.13 4±0.03 M⊙, R1 = 0.98±0.01 R⊙, M2 = 0.33 ±0.01 M⊙ and R2 = 0.92 ±0.01 R⊙ forV211; M1 = 1.304-0.05 M⊙, R1 = 1.03±0.01 R⊙, M2 = 0.58±0.02 M⊙ and R2 = 0.78±0.01 R⊙ for NV358. On the color-magnitude diagram of ω Centauri, V211 is located in the faint blue straggler region and its primary component is more massive than a star at the main-sequence turnoff. Therefore, V211 is a blue straggler and was formed by mass transfer from the secondary component to the primary. The age of NV358 is less than 1.93 Gyr, indicating that it is much younger than the first generation of stars in ω Centaufi. Like NV364 in ω Centaufi, NV358 should be a second-generation binary.

Simulation of old open clusters for UVIT on ASTROSAT

The Ultra Violet Imaging Telescope （UVIT） is one of the payloads on the first Indian multiwavelength satellite ASTROSAT, which is expected to be launched by the Indian Space Research Organisation （ISRO） in the year 2015. We have per- formed simulations of UV studies of old open clusters for the UVIT. The colour mag- nitude diagrams （CMDs） and spatial appearances have been created using 10 filters associated with the FUV channel （130-180 nm） and NUV channel （200-300 nm） that are available for observations on the UVIT, for the three old open clusters M67, NGC 188 and NGC 6791. The CMDs are simulated for different filter combinations, and they are used to identify the loci of various evolutionary sequences, white dwarfs, blue stragglers, red giants, subgiants, turn off stars and the main sequence of the clus- ters. The present work helps in identifying a potential area of study in the case of these three old open clusters by considering the availability of filters and the detection limits of the instrument. We also recommend filter combinations, which can be used to detect and study the above mentioned evolutionary stages. The simulations and the results presented here are essential for the optimal use of the UVIT for studies of old open clusters.

δScuti变星和相关天体

δ Scuti型变星是赫罗图上 A3～F5区间的主序及其以上的一种周期短于0.3 d的单周期或多周期小变幅脉动变星。与它们相关的变星有矮造父变星、γDol变星、蓝离散星、金属线星、A型特殊星、λ Boo变星和 δ Del变星。有些赫比格Ae/Be星也存在类似的脉动。对自 1964年起的研究工作进行了系统的总结,给出了1995年后新发现的相关变星数和最可靠的周期变化表,提出应当用双星轨道光时效应来解释实测得到的变星周期变化中的幅度很大的成分。统计表明自转越快变幅越小,因此年轻星团中不可能存在大变幅变星。变幅随周期的分布有3个极大值,最大变幅是周期0.17d处的1.0mag。恒星系统内变星的平均周期越短系统的年龄越大,金属丰度也越低。

Influence of the Choice of Core-Envelope Transition Point on the Binary Merger of Two Main-sequence Components

We have studied the influence of different choices of core-envelope transition point on the final merger of contact binaries with two main-sequence components. A binary of 1.00 + 0.90M⊙ with an initial orbital period of 0.35d is examined. The mass fraction of the primary mixed with the matter of the secondary, qmix, determined by the chosen core-envelope transition point, ranges from 0.04 to 1.00 in our analysis. If as qmix< 0.8, none of the helium-rich matter in the center of the primary is mixed into the envelope, and there is little distinction in the evolutionary tracks of the mergers. The timescales of the mergers remaining on the main sequence, tBS are very similar (-6.2×108yr) if qmix< 0.71, since no hydrogen-rich matter of the secondary is mixed into the core of the mergers; for qmix > 0.71, the larger qmix is, the greater the mixing, hence the longer the blue straggler lifetime, tBS, and also the greater the luminosity. For qmix= 1.00, tBS -8.5×108yr. Estimation by (?)r - (?)a = 0.0 shows that the point at which tBS begins to increase is about qmix= 0.68. In comparison with the homogeneously mixed models, the merger with a helium profile similar to that of the primary is less luminous and has a shorter tBS.

邱华栋小说游荡者形象的精神生态及启示

邱华栋是新生代作家的一个重要代表,其小说中游荡者系列形象的塑造是作家试图以边缘化的审美生存为现代都市人在商品与技术的牢笼中重新唤回情感与个性的一个尝试。游荡者们摆脱了生活必然性的限制,以审美的眼光悠游于都市,其生存挣扎及其价值追寻方式为我们探索新价值的重建提供了启示:边缘化为价值探索提供了合适的角度与场所,游荡所表征的现代审美意识为在理性化都市中重构日常生活结构提供了一种可行性。

论苏元春与会党的关系

苏元春早年参加过天地会张高友部,这段经历使他在以后的宦海生涯中比一般的文武官员更了解会党以及与之有密切联系的游勇,在实施剿抚兼施的策略时也往往更加有效。苏元春对陆荣廷等会党游勇剿抚兼施的处置策略是广西各种政治军事势力互相斗争、互相平衡以至互相妥协的结果,有着客观的合理性。它为边防建设提供了现成的兵员和劳力,减轻了边防的压力,有利于维持清政府在广西的统治。苏元春治军不严与他管辖的地方部队有关,他在十分困难的情况下,勉为其难,得过且过,部队时有失控,在所难免。说他"纵兵殃民"是言过其实。广西会党游勇泛滥成灾自有其外界的和自身的多种因素,其中中法战争的战后遗留问题,资产阶级各派在境外对本地会党的联络争取是广西特有的省情,苏元春一些做法不是也不可能是广西会党游勇问题的勃发的惟一因素。

蓝离散星研究现状(Ⅰ):形成机制

大量研究表明,蓝离散星的形成可能有多种机制,目前比较流行的形成机制可以概括为以下几类:密近双星系统的质量传输及双星并合、密集星场的恒星碰撞,以及包含双星系统的恒星间 (双星-单星、双星-双星)相互作用导致的恒星并合。与此同时,蓝离散星在各类恒星系统中的普遍存在,也使得研究这类恒星的形成及演化成为追踪恒星系统动力学演化、化学演化及积分光谱性质变化的有效指针。