期刊文献+
共找到6篇文章
< 1 >
每页显示 20 50 100
矿工不安全行为的网络传播性分析 被引量:23
1
作者 许正权 张妮 +1 位作者 王华清 曹庆仁 《科技进步与对策》 CSSCI 北大核心 2014年第11期54-56,共3页
煤矿日常生产中存在大量矿工不安全行为,矿工不安全行为一旦产生,在合适的条件下就有可能被其他矿工学习和模仿,并在特定的社会网络中传播。另外,在适当的条件下,不安全行为还会在传播过程中进行叠加,从而增加了其诱发煤矿事故的概率,... 煤矿日常生产中存在大量矿工不安全行为,矿工不安全行为一旦产生,在合适的条件下就有可能被其他矿工学习和模仿,并在特定的社会网络中传播。另外,在适当的条件下,不安全行为还会在传播过程中进行叠加,从而增加了其诱发煤矿事故的概率,导致矿工不安全行为的多发性、重复性及难预防性。构建了矿工行为状态空间模型,设定了矿工不同行为状态转换条件,并在此基础上对矿工不安全行为传播条件进行综合,指出行为的成本收益、行为的可学习和模仿价值及社会接触是影响矿工不安全行为传播性的主要条件,研究成果为干预矿工不安全行为提供了新的理论和方法支持。 展开更多
关键词 行为状态空间 社会网络 不安全行为 行为传播性
下载PDF
基于主智能体的群体学习算法GLBMA 被引量:4
2
作者 程显毅 李淑琴 夏德深 《江苏大学学报(自然科学版)》 EI CAS 北大核心 2005年第5期437-439,共3页
以机器人足球比赛(RoboCup)为背景,基于主智能体和辅助智能体概念,提出了基于主智能体群体强化学习算法(GLBMA),该算法通过主智能体和辅智能体的角色切换来实现整个团队的学习,改进了传统的群体强化学习算法。RoboCup仿真比赛试验表明,... 以机器人足球比赛(RoboCup)为背景,基于主智能体和辅助智能体概念,提出了基于主智能体群体强化学习算法(GLBMA),该算法通过主智能体和辅智能体的角色切换来实现整个团队的学习,改进了传统的群体强化学习算法。RoboCup仿真比赛试验表明,传统群体强化学习算法中的行为学习状态空间过大,连续状态空间的行为选择及多智能体合作求解等问题得到了解决. 展开更多
关键词 智能体 主智能体 强化学习 ROBOCUP 行为学习状态空间
下载PDF
多agent协同强化学习算法SE-MACOL及其应用 被引量:5
3
作者 王长缨 陈文伟 《广西师范大学学报(自然科学版)》 CAS 北大核心 2006年第4期167-170,共4页
针对多agent团队中各成员之间是协作关系且自主决策的学习模型,在此对Q学习算法进行了适当扩充,提出了适合于多agent团队的一种共享经验元组的多agent协同强化学习算法。其中采用新的状态行为的知识表示方法,采用相似性变换和经验元组... 针对多agent团队中各成员之间是协作关系且自主决策的学习模型,在此对Q学习算法进行了适当扩充,提出了适合于多agent团队的一种共享经验元组的多agent协同强化学习算法。其中采用新的状态行为的知识表示方法,采用相似性变换和经验元组的共享,能够提高多agent团队协同工作的效率。最后将该算法应用于猎人捕物问题域,实验结果表明该算法能够明显加快多个猎人合作抓捕猎物的进程。 展开更多
关键词 多AGENT学习 强化学习 Q学习 状态行为空间 协作团队
下载PDF
Some Dynamic and Combinatorial Properties of One Parameter Families of Unimodal Maps with Monotonicity
4
作者 John Taylor 《Journal of Mathematics and System Science》 2013年第6期301-308,共8页
It is known that certain one parameter families of unimodal maps of the interval have a topological universality with regard to their dynamic behavior [ 1, 2]. As a parameter is smoothly increased, a fascinating varie... It is known that certain one parameter families of unimodal maps of the interval have a topological universality with regard to their dynamic behavior [ 1, 2]. As a parameter is smoothly increased, a fascinating variety of dynamic behaviors are produced. For some families the behaviors are monotonic in the parameter, while in others they are not [3]. The question is what sort of conditions on a one parameter family will ensure this monotonicity of the behavior with the parameter? The answer is unknown and will not be given here. What we do instead is to investigate certain geometric-dynamic-combinatorial consequences of assuming that the family has this monotonicity. Specifically, using tools of symbolic dynamics, state space is "course grained" with a finite alphabet. We decompose a non-invertible map into nonlinear but invertible pieces. From these invertible pieces, we form inverse maps via composition along words. Equations of motion are developed for both forward and inverse orbits (in both the variables of state space and the parameter), and an equation relating forward and inverse motions at fix-points is exhibited. Finally, we deduce a list of conditions, each of which is equivalent to monotone behavior. One of these conditions states that simple parity characteristics of words correspond to definite dynamics near fixed-points and vice versa. 展开更多
关键词 One parameter family unimodal map kneading theory connection equation.
下载PDF
一种共享经验元组的多agent协同强化学习算法 被引量:4
5
作者 王长缨 尹晓虎 +1 位作者 鲍翊平 姚莉 《模式识别与人工智能》 EI CSCD 北大核心 2005年第2期234-239,共6页
Q学习算法是一种最受欢迎的模型无关强化学习算法。本文通过对Q学习算法进行合适的扩充,提出了一种适合于多agent协作团队的共享经验元组的多agent协同强化学习算法,其中采用一种新的状态行为的知识表示方法使得状态行为空间得到缩减,... Q学习算法是一种最受欢迎的模型无关强化学习算法。本文通过对Q学习算法进行合适的扩充,提出了一种适合于多agent协作团队的共享经验元组的多agent协同强化学习算法,其中采用一种新的状态行为的知识表示方法使得状态行为空间得到缩减,采用相似性变换和经验元组的共享使得学习的效率得到提高。最后将该算法应用于猎人捕物问题域。实验结果表明该算法能够加快多个猎人合作抓捕猎物的进程,有利于协作任务的成功执行,并能提高多agent协作团队的协作效率,因此该算法是有效的。 展开更多
关键词 多AGENT学习 强化学习 Q学习 状态行为空间 协作团队
原文传递
Three-dimensional dissipative stress space considering yield behavior in deviatoric plane 被引量:4
6
作者 YAO YangPing CUI WenJie WANG NaiDong 《Science China(Technological Sciences)》 SCIE EI CAS 2013年第8期1999-2009,共11页
Naturally deposited soils are always found in the complex three-dimensional stress state.Constitutive models developed for modeling the three-dimensional mechanical behavior of soils should obey the basic laws of ther... Naturally deposited soils are always found in the complex three-dimensional stress state.Constitutive models developed for modeling the three-dimensional mechanical behavior of soils should obey the basic laws of thermo-mechanical principles.Based on the incremental dissipation function,a new deviatoric shift stress is derived and then introduced into the existing constitutive models to describe the yield behavior in the deviatoric plane for geomaterials.By adopting the proposed shift stress,the relationship between dissipative stress tensors and true stress tensors can be established.Therefore,the threedimensional plastic strain can be calculated reasonably through the associated flow rule in the three-dimensional dissipative stress space.At the same time,three methods that are conventionally adopted for generalizing constitutive models to model the three-dimensional stress-strain relationships are examined under the thermo-mechanical framework.The TS(transformed stress)method is shown to obey the thermo-mechanical rules and the TS space adopted in TS method is actually a translational three-dimensional dissipative stress space.However,it is illustrated that the other two approaches,the method of using failure criterion directly and the method of using g()function,violate the basic rules of thermo-mechanical theories although they may bring convenience and simplicity to numerical analysis for geotechnical engineering.Comparison between model predictions and experimental data confirms the validity of the proposed three-dimensional dissipative stress space. 展开更多
关键词 SOILS thermo-mechanics yield behavior shift stress stress space THREE-DIMENSION
原文传递
上一页 1 下一页 到第
使用帮助 返回顶部