从信息学的角度分析复杂网络链路预测

链路预测作为复杂网络分析的一项重要任务,其目的是寻找节点间缺失(新)的链路,识别虚假交互,对于挖掘和分析网络的演化,重塑网络模型具有重要意义.传统的链路预测方法多数采用拓扑结构信息、节点的属性信息和图的结构特征.应用这些特征等外部信息可以得到很好的预测效果.本文从信息学的角度全面分析、回顾和讨论了复杂网络链路预测的发展现状,提出了链路预测技术和问题的系统分类.首次将分层的思想引入链路预测分类体系中,把当前的链路预测方法分为基于监督学习的技术、基于半监督学习的技术、基于无监督学习的技术和基于强化学习的技术.对每种技术的优缺点、复杂性、所使用的具体特征,开源实现及应用建议进行了详细的分析.最后,讨论了当前复杂网络链路预测技术未来的发展方向.     

移动机器人运动规划中的深度强化学习方法

随着移动机器人作业环境复杂度的提高、随机性的增强、信息量的减少,移动机器人的运动规划能力受到了严峻的挑战.研究移动机器人高效自主的运动规划理论与方法,使其在长期任务中始终保持良好的复杂环境适应能力,对保障工作安全和提升任务效率具有重要意义.对此,从移动机器人运动规划典型应用出发,重点综述了更加适应于机器人动态复杂环境的运动规划方法——深度强化学习方法.分别从基于价值、基于策略和基于行动者-评论家三类强化学习运动规划方法入手,深入分析深度强化学习规划方法的特点和实际应用场景,对比了它们的优势和不足.进而对此类算法的改进和优化方向进行分类归纳,提出了目前深度强化学习运动规划方法所面临的挑战和亟待解决的问题,并展望了未来的发展方向,为机器人智能化的发展提供参考.     

中央空调系统的 MAS 智能控制模型

中央空调智能控制是实现空调节能优化调节的一个发展方向,目前是研究的一个热点和难点.文中首先在分布式人工智能领域的 MAS 理论基础上,对 MAS 的理论和应用进行研究介绍;然后针对目前中央空调控制系统中难于进行整体控制的问题,采用基于 MAS 的建模方法进行整体控制的设想,探讨了多 Agent 技术在空调控制系统中的应用;最后,文中引入了具有学习控制能力的人工智能技术,采用增强学习方法与多 Agent 技术结合起来,实现中央空调系统的整体智能控制与节能     

深度强化学习研究综述

深度强化学习是指利用深度神经网络的特征表示能力对强化学习的状态、动作、价值等函数进行拟合,以提升强化学习模型性能,广泛应用于电子游戏、机械控制、推荐系统、金融投资等领域。回顾深度强化学习方法的主要发展历程,根据当前研究目标对深度强化学习方法进行分类,分析与讨论高维状态动作空间任务上的算法收敛、复杂应用场景下的算法样本效率提高、奖励函数稀疏或无明确定义情况下的算法探索以及多任务场景下的算法泛化性能增强问题,总结与归纳4类深度强化学习方法的研究现状,同时针对深度强化学习技术的未来发展方向进行展望。     

中央空调系统的MAS智能控制模型

中央空调智能控制是实现空词节能优化调节的一个发展方向,目前是研究的一个热点和难点。文中首先在分布式人工智能领域的MAS理论基础上,对MAS的理论和应用进行研究介绍;然后针对目前中央空调控制系统中难于进行整体控制的问题,采用基于MAS的建模方法进行整体控制的设想,探讨了多Agent技术在空调控制系统中的应用;最后,文中引入了具有学习控制能力的人工智能技术,采用增强学习方法与多Agent技术结合起来,实现中央空调系统的整体智能控制与节能。     

深度学习在故障诊断领域中的研究现状与挑战

现代工业系统已呈现出向大型化、复杂化的方向发展,使得针对工业系统的故障诊断方法遇到一系列的技术难题.近年来,深度学习(deep learning)在特征提取与模式识别方面显示出独特的优势与潜力,将深度学习应用于解决复杂工业系统故障诊断的研究已初现端倪.为此,首先介绍几种典型的基于深度学习方法实现工业系统故障诊断方法;然后对基于深度学习实现故障诊断的主要思想和建模方法进行描述;最后总结和讨论了复杂工业系统故障的特点,并探讨了深度学习在实现复杂工业系统故障诊断方面所面临的挑战,展望了未来值得继续研究的方向.     

强化学习的自动驾驶控制技术研究进展

自动驾驶车辆的本质是轮式移动机器人,是一个集模式识别、环境感知、规划决策和智能控制等功能于一体的综合系统。人工智能和机器学习领域的进步极大推动了自动驾驶技术的发展。当前主流的机器学习方法分为：监督学习、非监督学习和强化学习3种。强化学习方法更适用于复杂交通场景下自动驾驶系统决策和控制的智能处理,有利于提高自动驾驶的舒适性和安全性。深度学习和强化学习相结合产生的深度强化学习方法成为机器学习领域中的热门研究方向。首先对自动驾驶技术、强化学习方法以及自动驾驶控制架构进行简要介绍,并阐述了强化学习方法的基本原理和研究现状。随后重点阐述了强化学习方法在自动驾驶控制领域的研究历史和现状,并结合北京联合大学智能车研究团队的研究和测试工作介绍了典型的基于强化学习的自动驾驶控制技术应用,讨论了深度强化学习的潜力。最后提出了强化学习方法在自动驾驶控制领域研究和应用时遇到的困难和挑战,包括真实环境下自动驾驶安全性、多智能体强化学习和符合人类驾驶特性的奖励函数设计等。研究有助于深入了解强化学习方法在自动驾驶控制方面的优势和局限性,在应用中也可作为自动驾驶控制系统的设计参考。     

迭代学习控制的研究现状

迭代学习控制经历了二十多年的发展历程,已经取得了很多研究成果,现已成为智能控制的一个重要研究方向,并得到越来越广泛的应用.本文对迭代学习控制的基本原理和主要研究问题从发展的角度作了详细阐述,并对其应用作了细致介绍.     

LoRa技术在数字农业中的应用

从国内外发展来看,物联网技术正驱动农业向数字化、精细化方向发展。LoRa技术具备广域覆盖、低功耗、低成本和抗干扰能力强等特性。农业环境较复杂,需要传感器定期上传数据,且WiFi、ZigBee、蓝牙技术不能兼顾远距离和低功耗。基于此,以农业大棚为例,提出采用LoRa技术进行通信,将传感器采集的环境数据传输至远程或者本地服务器,实现农田环境管理。     

智能控制：从学习控制到平行控制

20世纪60年代,学习控制开启了人类探究复杂系统控制的新途径,基于人工智能技术的智能控制随之兴起.本文以智能控制为主线,阐述其由学习控制向平行控制发展的历程.本文首先介绍学习控制的基本思想,描述了智能机器的架构设计与运行机理.随着信息科技的进步,基于数据的计算智能方法随之出现.对此,本文进一步简述了基于计算智能的学习控制方法,并以自适应动态规划方法为切入点分析非线性动态系统自学习优化问题的求解过程.最后,针对工程复杂性与社会复杂性互相耦合的复杂系统控制问题,阐述了基于平行控制的学习与优化方法求解思路,分析其在求解复杂系统优化控制问题方面的优势.智能控制思想经历了学习控制、计算智能控制到平行控制的演化过程,可以看出平行控制是实现复杂系统知识自动化的有效方法.     

Incremental online sparsification for model learning in real-time robot control

For many applications such as compliant, accurate robot tracking control, dynamics models learned from data can help to achieve both compliant control performance as well as high tracking quality. Online learning of these dynamics models allows the robot controller to adapt itself to changes in the dynamics (e.g., due to time-variant nonlinearities or unforeseen loads). However, online learning in real-time applications - as required in control - cannot be realized by straightforward usage of off-the-shelf machine learning methods such as Gaussian process regression or support vector regression. In this paper, we propose a framework for online, incremental sparsification with a fixed budget designed for fast real-time model learning. The proposed approach employs a sparsification method based on an independence measure. In combination with an incremental learning approach such as incremental Gaussian process regression, we obtain a model approximation method which is applicable in real-time online learning. It exhibits competitive learning accuracy when compared with standard regression techniques. Implementation on a real Barrett WAM robot demonstrates the applicability of the approach in real-time online model learning for real world systems.     

深度学习在软件定义网络研究中的应用综述

数据转发与控制分离的软件定义网络（Software Defined Networking,简称SDN）是对传统网络架构的彻底颠覆,为网络各方面的研究引入新的机遇和挑战.随着传统网络研究方法在SDN中遭遇瓶颈,基于深度学习的方法被引入到SDN的研究中,在实现实时智能的网络管控上成果颇丰,推动了SDN研究的深入发展.调查了深度学习开发平台,训练数据集,智能SDN架构等深度学习引入SDN的促进因素;对智能路由,入侵检测,流量感知和其他应用等SDN研究领域中的深度学习应用进行系统的介绍,深入分析了现有深度学习应用的特点和不足;最后展望了SDN未来的研究方向与趋势.     

Ballistic learning control： formulation, analysis and convergence

In this paper, we formulate and explore the characteristics of iterative learning in ballistic control problems. The iterative learning control (ILC) theory provides a suitable framework for derivations and analysis of ballistic control under learning process. To overcome the obstacles caused by uncertain gradient and redundant control input, we incorporate extra trials into iterative learning. With the help of trial results, proper control and updating direction can be determined. Then, iterative learning can be applied to ballistic control problem. Several initial state learning algorithms are studied for initial speed control, force control, as well as combined speed and angle control. In the end, shooting angle learning in the basketball shot process is simulated to verify the effectiveness of iterative learning methods in ballistic control problems.     

A study of overfitting in optimization of a manufacturing quality control procedure

Quality control of the commutator manufacturing process can be automated by means of a machine learning model that can predict the quality of commutators as they are being manufactured. Such a model can be constructed by combining machine vision, machine learning and evolutionary optimization techniques. In this procedure, optimization is used to minimize the model error, which is estimated using single cross-validation. This work exposes the overfitting that emerges in such optimization. Overfitting is shown for three machine learning methods with different sensitivity to it (trees, additionally pruned trees and random forests) and assessed in two ways (repeated cross-validation and validation on a set of unseen instances). Results on two distinct quality control problems show that optimization amplifies overfitting, i.e., the single cross-validation error estimate for the optimized models is overly optimistic. Nevertheless, minimization of the error estimate by single cross-validation in general results in minimization of the other error estimates as well, showing that optimization is indeed beneficial in this context.     

Case studies in high-dimensional classification

We consider the application of several compute-intensive classification techniques to two significant real-world applications: disk drive manufacturing quality control and the prediction of chronic problems in large-scale communication networks. These applications are characterized by very high dimensions, with hundreds of features or tens of thousands of cases. The results of several learning techniques are compared, including linear discriminants, nearest-neighbor methods, decision rules, decision trees, and neural nets. Both applications described in this article are good candidates for rule-based solutions because humans currently resolve these problems, and explanations are critical to determining the causes of faults. While several learning techniques achieved competitive results, machine learning with decision rule inducton was most effective for these applications. It is demonstrated that decision (production) rule induction is practical in high dimensions, providing strong results and insightful explanations.This research was performed while the author was a visiting researcher at IBM T.J. Watson Research Center and AT&T Bell Labs.     

LEARNING TO SOLVE PROBLEMS FROM EXERCISES

It is a common observation that learning easier skills makes it possible to learn the more difficult skills. This fact is routinely exploited by parents, teachers, textbook writers, and coaches. From driving, to music, to science, there hardly exists a complex skill that is not learned by gradations. Natarajan's model of “learning from exercises” captures this kind of learning of efficient problem solving skills using practice problems or exercises ( Natarajan 1989 ). The exercises are intermediate subproblems that occur in solving the main problems and span all levels of difficulty. The learner iteratively bootstraps what is learned from simpler exercises to generalize techniques for solving more complex exercises. In this paper, we extend Natarajan's framework to the problem reduction setting where problems are solved by reducing them to simpler problems. We theoretically characterize the conditions under which efficient learning from exercises is possible. We demonstrate the generality of our framework with successful implementations in the Eight Puzzle, symbolic integration, and simulated robot planning domains illustrating three different representations of control knowledge, namely, macro‐operators, control rules, and decision lists. The results show that the learning rates for the exercises framework are competitive with those for learning from problems solved by the teacher.     

A structured matrix approach to efficient calculation of LQG repetitive learning controllers in the lifted setting

This article discusses the use of repetitive control for output reference tracking in linear time-varying discrete time systems with both repetitive and non-repetitive noise components. The design of such controllers is formulated as a lifted linear stochastic output feedback problem on which the mature techniques of discrete linear control may be applied. In many modern applications, the large size of the system matrices in such a control problem inhibits the application of standard solvers and optimisation techniques. For linear quadratic Gaussian (LQG) problems, the matrices of the lifted feedback problem can be fitted into the recently developed sequentially semi-separable structure. Innovative numerical solutions are developed that have 𝒪(N) computational complexity (where N is the trial length) in both controller synthesis and implementation, comparable to that of many non-lifted and Fourier transform based learning control methods. Moreover, within this formulation, the system is allowed to vary over the learning cycle, closed-loop stability is guaranteed, and stochastic noise and disturbances are handled in an LQG sense.     

多智能体强化学习在城市交通网络信号控制方法中应用综述

交通信号控制系统在物理位置和控制逻辑上分散于动态变化的网络交通环境,将每个路口的交通信号控制器看作一个异质的智能体,非常适合采用“无模型、自学习、数据驱动”的多智能体强化学习（MARL）方法建模与描述。为了解该方法的研究现状、存在问题及发展前景,系统跟踪了多智能体强化学习在国内外交通控制领域的具体应用,包括交通信号MARL控制概念模型、完全孤立的MARL控制、部分状态合作的MARL和动作联动的MARL控制,分析其技术特征和代际差异,讨论了多智体强化学习方法在交通信号控制中的研究动向,提出了发展网络交通信号多智能体强化学习集成控制的关键问题在于强化学习控制机理、联动协调性、交通状态特征抽取和多模式整合控制。     

深度强化学习在智能制造中的应用展望综述

深度强化学习作为机器学习发展的最新成果,已经在很多应用领域崭露头角。关于深度强化学习的算法研究和应用研究,产生了很多经典的算法和典型应用领域。深度强化学习应用在智能制造中,能在复杂环境中实现高水平控制。对深度强化学习的研究进行概述,对深度强化学习基本原理进行介绍,包括深度学习和强化学习。介绍深度强化学习算法应用的理论方法,在此基础对深度强化学习的算法进行了分类介绍,分别介绍了基于值函数和基于策略梯度的强化学习算法,列举了这两类算法的主要发展成果,以及其他相关研究成果。对深度强化学习在智能制造的典型应用进行分类分析。对深度强化学习存在的问题和未来发展方向进行了讨论。