移动边缘计算中的在线任务卸载方法

为减少移动边缘计算(mobile edge computing,MEC)网络中移动用户的长期任务开销,利用强化学习的马尔科夫决策过程,将用户的移动性与系统的动态信息建模为随机优化问题.依据系统信息的状态,将问题分为系统信息已知、系统信息未知2种情况.在系统信息已知时,提供了问题的最优解;系统信息未知时,基于在线学习提出...     

具有一个吸收态的有限非时齐马氏链的PH分布

首先证明了有限非时齐马氏链的转移概率矩阵满足C-K方程,然后给出了具有一个吸收态的有限非时齐马氏链的PH分布.     

分层强化学习原理研究

首先介绍了强化学习基本原理,分析了马尔科夫决策过程与半马尔科夫决策过程的理论基础及其在强化学习中的应用,其次阐述了分层强化学习中分层与抽象的思想,分析了HAM、Options与MaxQ等方法,并从分层与抽象角度进行了比较,最后指出了分层强化学习的研究发展方向。     

A study on forecasting electricity production and consumption in smart cities and factories

The electrical power sector must undergo a thorough metamorphosis to achieve the ambitious targets in greenhouse gas reduction set forth in the Paris Agreement of 2015. Reducing uncertainty about demand and, in case of renewable electricity generation, supply is important for the determination of spot electricity prices. In this work we propose and evaluate a context-based technique to anticipate the electricity production and consumption in buildings. We focus on a household with photovoltaics and energy storage system. We analyze the efficiency of Markov chains, stride predictors and also their combination into a hybrid predictor in modelling the evolution of electricity production and consumption. All these methods anticipate electric power based on previous values. The main goal is to determine the best method and its optimal configuration which can be integrated into a (possibly hardware-based) intelligent energy management system. The role of such a system is to adjust and synchronize through prediction the electricity consumption and production in order to increase self-consumption, reducing thus the pressure over the power grid. The experiments performed on datasets collected from a real system show that the best evaluated predictor is the Markov chain configured with an electric power history of 100 values, a context of one electric power value and the interval size of 1.     

基于HMM状态结构调整的非特定人语音识别

利用HMM模型状态间的混淆度, 提出了一种新的状态结构调整算法, 使不同的状态可以共享相同的高斯混合函数, 并在EM算法的框架下推导出对状态结构调整后的增加参数, 即状态间权值的重估公式. 并对非特定人进行大词汇量汉语连续语音识别实验, 实验结果表明状态结构调整后的系统不仅优于基线系统, 还获得了比传统的参数增加方法更高的识别率, 由此证明了状态结构调整方法的有效性.     

一类马氏过程的大偏差

讨论了{（X^ε（t）,Z^ε（t））;ε〉0,t∈[0,T]}的大偏差性质,X^ε（t）满足下面的随机微分方程： dX^ε（t）=√εσ（X^ε（t））dB（t）＋b（X^ε（t）,Z^ε（t））dt Z^ε（t）为有限个状态的随机过程．     

概率方法在解决实际问题中的应用

举例说明如何利用概率论和随机过程的相关知识建立目标函数,求出最优值来解决实际问题。     

A cross-layer design approach on spectrum allocation and resource scheduling in cognitive PMP networks

We propose the spectrum allocation and resource scheduling algorithms in cognitive point to multipoint （PMP） networks with rapid changes of spectrum opportunities and present a media access control （MAC） protocol based on these algorithms. The objective of spectrum allocation is to make efficient use of the spectrum while maintaining the transceiver synchronization on frequency and time in the network. The objective of resource scheduling is to guarantee the quality of service （QoS） requirements of different kinds of connections and to minimize the total energy consumption in the network as well. By sensing only a small set of possible channels in each slot based on the state transition probability of each channel, our spectrum allocation algorithm achieves high spectrum efficiency in the network. The resource scheduling problem is divided into three sub problems and we derive optimal solutions to these problems by greedy algorithm and convex optimization. The simulation results show that our algorithm can make efficient use of the spectrum and the network resources at a cost of low computational complexity.     

大数据时代大学生自主学习能力预测研究

大数据时代知识增长速度呈指数级增长,提升大学生自主学习能力迫在眉睫。根据大学生自主学习能力调查问卷,利用马尔可夫链构建大学生自主学习能力预测模型,通过模糊综合评判法对大学生自主学习能力进行评估,并对实验结果进行分析比较,给出不同年级、不同性别以及学习动机对自主学习能力的影响,最后给出大学生提升自主学习能力的方法与建议。     

遗传算法的收敛性研究

通过马尔可夫链方法,分析种群在解空间上概率分布情况以及收敛到最优解的概率,证明经典GA是不会收敛到最优解的,若在GA中保留每一代的最佳个体,则可以收敛到最优解。讨论全局收敛和过早收敛的原因,最后提出GA操作中应遵循的原则是改进GA搜索性能的关键。