AI算法在车联网通信与计算中的应用综述

在5G时代，车联网的通信和计算发展受到信息量急速增加的限制。将AI算法应用在车联网，可以实现车联网通信和计算方面的新突破。调研了AI算法在通信安全、通信资源分配、计算资源分配、任务卸载决策、服务器部署、通算融合等方面的应用，分析了目前AI算法在不同场景下所取得的成果和存在的不足，结合车联网发展趋势，讨论了AI算法在车联网应用中的未来研究方向。     

异构计算带来AI视觉新突破

近年来,随着图像处理及识别等关键研究领域的突破,使人工智能在视觉应用上获得了前所未有的进步,成为AI大规模应用推广的第一个场景.为此,笔者采访了华夏芯(北京)通用处理器技术有限公司首席执行官李科奕先生,并就相关内容进行了解答.     

芯片厂商引领广电的人工智能化

本文从芯片厂商的角度分析人工智能(AI)得以迅速发展的三大关键要素.并以当前AI中最核心的深度学习技术为例,分析其训练和执行两个阶段分别对硬件的需求,得出GPU和CPU性能在AI应用的重要性.随后介绍了兆芯在GPU和CPU方面的技术优势,以及在广电领域,兆芯凭借自身平台优势引领AI产业发展及推进AI应用落地的实际案例.     

星环科技与索为科技达成战略合作

近日,星环科技与索为科技正式签约为战略合作伙伴,共同探索数据中台和人工智能领域的应用,达成多行业、多场景的深度合作。本次合作将围绕数据中台生态建设、数据应用场景实践、AI技术落地和应用示范等方面展开合作。此次合作,星环科技将与索为科技共同搭建多个数据应用场景,索为科技借助星环科技在大数据、人工智能、云计算、数据安全领域的领先技术和产品落地能力,共同推动数据驱动的业务创新。     

AI在5G行业专网的应用场景和使能技术

行业专网是5G网络的重要组成部分,电信运营商通过AI与行业专网应用场景结合实现业务SLA保障和网络优化.面向数字化、智能化的新一代行业专网发展目标,基于行业用户的体验保障和网络安全需求探讨了AI在5G行业专网典型应用场景,分析了基于AI的5G行业专网应用的关键使能技术,包括算力网络、自适应优化AI算法、意图驱动的网络织...     

网络人工智能发展分析与建议

人工智能(AI)技术经历了60多年的发展并达到前所未有的高度,通信网络领域是其最大的细分市场。简要介绍了AI技术的发展历程,以及本轮AI技术发展的内涵和核心要素。通过AI在通信网络领域应用发展的驱动力和概况的阐述,提出5大典型网络AI应用场景,并针对3大核心要素给出网络AI的发展建议。     

智慧视觉的算法、研发和产品动向

人工智能(AI)开始由科研走向行业应用,其中嵌入式视觉是重要的应用场景,例如汽车、监控、虚拟现实/增强现实、医学诊断、工业视觉、无人机和移动市场等.为此,笔者探访了业内部分相关企业,请他们介绍了部分算法、研发和产品动向.     

内生智能和端到端服务化的6G无线网络架构设计

为了解决5G网络能耗高、结构复杂和运维管理难度大等问题,同时使得网络具备更极致的全场景适应能力,从网络功能、网络结构、网络运行三个方面对未来6G网络进行设计,实现6G网络的智慧内生和极简。在网络功能层面,提出端到端微服务化的架构,通过按需功能组合与编排提升行业用户乃至个人用户定制化需求的满足能力;在网络结构层面,提出一种低、中、高多频段协同的信令广域覆盖机制及相关方案设计,实现多制式、多频段小区的统一管理,进一步降低整网信令开销、小区管理复杂度和网络能耗;在网络运营层面,不同于5G外挂式AI,提出网络内生智能设计和AI服务质量(Quality of AI Service,QoAIS)指标体系及保障机制,将AI打造成6G网络的一种基础能力和服务。最后,对内生AI网络架构中基于QoAIS的AI工作流编排方案进行了仿真分析。仿真结果表明,在性能和开销方面,三种典型方案(集中式、分布式和协同式)各有优缺点,需要针对QoAIS的具体要求,决定AI工作流中相关任务的分布,并调度所需资源,以保证QoAIS的达成。     

5G自身智能化及赋能智能产业之路

2019年是5G商用的元年,5G深入各个垂直行业的应用同人工智能(AI)的典型应用高度重叠。5G网络自身的智能化是5G+AI赋能各行各业应用的基础,同时也是提升网络效率,降低运营运维成本的关键。5G+AI赋能产业的场景和规模正在加速发展,这些不断拓展的新业务需求也对5G网络和AI技术提出了新的挑战。本文阐述了5G网络自身智能化及5G+AI赋能产业所取得的进展,以及面临的核心技术问题,呼吁产学研突破现有的研发共享合作模式,加速解决核心技术挑战。     

高通徐晧：“5G+AI”双擎赋能 激发数字经济创新活力

<正>5G与AI的融合发展是推动数字经济向智能经济延伸升级的关键，随着我国数字经济发展步入“快车道”，AI使能5G、5G“反哺”AI的现象频现，二者相得益彰，智慧赋能千行百业。在2023世界电信和信息社会日大会上，高通中国区研发负责人徐晧表示，高通正把握“5G+AI”协同发展带来的战略机遇，持续推动5G和AI技术演进，打造领先产品，拓展行业新应用、新场景，在开启智能手机全新创新周期的同时，加速千行百业数字化转型。     

Artificial intelligence in power system operations

Power system operators often reach a cognitive barrier when information arrives too fast during a power system emergency. At such times it becomes difficult to reach a correct diagnosis of the problem or to formulate the correct decision when actions must be taken. Artificial Intelligence gives designers of Energy Management Systems a way to solve many of the diagnosis and decision problems so as to make the EMS more useful. This paper explores reasons why AI techniques, such as knowledge-based expert systems, are being used in EMS designs and the differences between knowledge-based expert systems and traditional numeric algorithm development. The differences between expert systems and the numeric approach extend to the basic conception and design of the applications. This is illustrated using a relay fault diagnosis system, showing both the traditional and rapid prototyping approaches to its development. Finally, issues concerned with the implementation of AI in EMS computers are explored along with the authors' predictions of possible AI applications to power system operations.     

How a Convergence Product Affects Related Markets: The Case of the Mobile Phone

Analyzing the diffusion of a convergence product is a new and challenging research field. It is very difficult to find research dealing with this issue due to the inherent complexity and lack of data. In analyzing the diffusion of a convergence product, we should simultaneously take into account its relationship with related single‐function products because of their similarities in terms of technology and functionality. In this study, we empirically analyze the diffusion of the convergence mobile phones in South Korea and find that the convergence products can affect the diffusion of MP3 players and digital cameras positively or negatively. This research may be significant for business strategies in technology management and product development.     

双层双联微模块设计方案

随着5G时代的到来及物联网、云计算、AI等新兴技术的迅速发展,数据业务形成了大容量、大带宽、大连接、低时延等新特征,这就要求数据中心能够实现大规模、快速灵活部署和弹性扩展。传统数据中心建设模式面临的主要问题：建设周期长,土建建设与业务需求不匹配;业务部署灵活性、扩展性不佳。基于上述矛盾,具备装配化、模块化特征的微型模块化数据中心解决方案应用而生。本文以双层双联微模块方案研究为例,总结分享新技术研究历程,为后续产品化设计提供借鉴。     

基于AI深度学习的面向业务5G基站节能系统研究

由于5G业务发展,5G基站数量增多,造成运营商的电费成本急剧增加,节能降耗成为运营商的可持续发展需求。在研究主流5G基站节能模式及多方位5G节能方案的基础上,提出了基于多种AI算法的5G基站节能系统,通过单SIM卡级别的高精度业务识别,在保证5G重要业务等各类型业务稳定运行的基础上,实现了最佳策略的5G基站柔性节能。     

Business Models and Business Model Innovation in a “Secure and Distributed Cloud Clustering (DISC) Society”

The development and innovation of business models to a secure distributed cloud clustering society (DISC)—is indeed still a complex venture and has not been widely researched yet. Numerous types of security technologies are in these years proposed and in the “slip stream” of these the study of secure business models and how business models can be operated and innovated in a secure context have intensified tremendously. The development of new mobile and wireless security technologies gives hopes to really realize a secure cloud clustering society where business models can act and be innovated secure—but we still have some steps to go before we reach the final destination. The paper gives a conceptual futuristic outlook on behalf of the input from SW2010 and state of the art business model research to what we can expect of business Model and business model innovation in a future secure cloud clustering society (DISC).     

The R&D cycle: the influence of product and process R&D onshort-term ROI

The purposes of this paper are first to explore the relationship between the total R&D budget and two major components, product and process R&D, and second, to examine the relationship between these two types of R&D and the profitability of the business. The paper explores a cyclical pattern in the relationship between product and process R&D on the one hand and the short-term return on investment (ROI) of the business on the other hand. The total level of R&D investment plays a key role in this relationship. The foundation of this paper is based on prior work about the changing role of R&D over the life cycle of a new product or a new technology. It is also based on the notions of the S-curve response function and the limited compatibility between product and process R&D. The empirical part of the paper is derived from the PIMS database where the sampling units are strategic business units (SBUs). It is, therefore, different from studies where the focus of analysis is a single product or technology. Two forms of analysis are presented: correlation analysis of the hypothesized relationship and nonparametric tests of the stability of findings. The analysis assumes a lagged relationship between the investment in R&D and performance in terms of ROI. The primary conclusion is that product and process R&D modestly affect ROI two years later. The relationship appears to be different depending on the level of total R&D spending     

Survivable virtual topology mapping in IP-over-WDM networks using differential evolution optimization

In IP-over-wavelength division multiplexing networks, a virtual topology is placed over the physical topology of the optical network. Given that a simple link failure or a node failure on the physical topology can cause a significant loss of information, an important challenge is to make the routing of the virtual topology on to the physical topology survivable. This problem is known as survivable virtual topology mapping (SVTM) and is known to be an NP-complete problem. So far, this problem has been optimally solved for small instances by the application of integer linear programming and has been sub-optimally solved for more realistic instances by heuristic strategies such as ant colony optimization and genetic algorithms. In this paper, we introduce the application of differential evolution (DE) to solve the SVTM problem and enhancements based on DE are proposed as well. Three algorithms based on DE are developed. The enhanced variants have better convergence rate, get better quality of solutions and require few control parameters. We present the impact of these parameters on the system’s performance improvement. Algorithms are evaluated in different test bench optical networks, as NSFnet and USA, demonstrating that the enhanced DE algorithm overcomes the other two, for small instances. The three algorithms reach a 100  survivable mapping for small instances. The three algorithms also find positive survivable mappings and reduce the network wavelength links. Results show the effectiveness and efficiency of the proposed algorithms.     

A prescriptive production-distribution approach for decision makingin new product design

The production problem in product design consists of determining technically feasible options and implementing collective decision making to decide which alternatives to produce. A problem with considering collective decision making in the production problem is that it may not be possible to define, a priori, a group utility function, due to the difficulty of making interpersonal comparisons of the members' preferences. Constructing a group utility function and solving the production problem may lead to unacceptable alternatives for some of the members. The distribution problem involves consideration of individual preferences in satisfying group decision-making situations. If production and distribution of the products are considered simultaneously, the preferences of the members are taken into account, explicitly leading to acceptable solutions. This approach also allows for explicit consideration of a company business strategy in determining which products to develop. A production-distribution approach is outlined and demonstrated for the design of a power electronics module product with a group of decision makers consisting of three customers and a manufacturer. In addition to explicit consideration of the customers' preference, some business strategies are also considered, leading to product alternatives acceptable to the customers and manufacturer     

AI大数据在审计管理中的应用研究

通过对业务需求和技术现状的分析，文中深入研究了各种审计模型和关键业务流程，并运用大数据平台建立的算法分析模型，从海量文件和票据中快速定位风险。集成了AI基础能力的系统，可以构建以AI技术实现对业务场景的服务支撑能力，实现AI技术在智慧审计领域的应用落地，为审计业务提供智能化支撑，丰富精细化管理手段的同时减少人工成本，实现降本增效，提升审计人员的工作效率。