1) decision gradient
决策梯度
2) policy gradient
策略梯度
1.
Theories, Algortihms and Applications of Policy Gradient Reinforcement Learning;
策略梯度增强学习的理论、算法及应用研究
2.
The adaptive heuristic critic(AHC) reinforcement learning frame is approximate of the value function and the policy function of Markov decision process(MDP),the stochastic MDPs can be converted to deterministic MDPs by the policy gradient reinforcement learning.
自适应启发评价(AHC)增强学习结构分别逼近马尔可夫决策过程的值函数和策略函数,策略梯度增强学习能够将随机不确定的马尔可夫决策过程转换为确定性的马尔可夫决策过程。
3.
Although policy gradient reinforcement learning (PGRL) has good convergence properties, the variance of policy gradient estimation in existing PGRL algorithms is usually large, which becomes a significant problem for policy gradient algorithms in theory and in practice.
尽管策略梯度强化学习算法有较好的收敛性,但是在梯度估计的过程中方差过大,却是该方法在理论和应用上的一个主要弱点。
4) amplitude decision
幅度决策
5) dispatching decision-making
调度决策
1.
Using the theory of system optimization and management and based on summarizing and analyzing the domestic and abroad research findings, relevant problems confronted in the course of dispatching decision-making for water supply were studied deep.
应用系统优化与管理学的理论,在总结分析国内外研究成果的基础上,对城市供水运行调度决策中涉及到的各个环节问题进行了深入研究。
6) decision power
决策强度
1.
The reasons caused the disadvantages lie in that the two formulas for measuring mean decision power and decision e.
通过实例分析,指出两种新的属性约简概念的不足,其根源是度量平均决策强度和决策熵的两个公式不具有单调性。
2.
Using the support and confidence of the rule in data mining,the decision power of the decision table is defined.
提出了决策表的决策强度的概念,分析了在属性约简过程中,决策强度的变化趋势;然后给出一种基于决策强度的属性约简的启发式算法,该算法的时间复杂度为 O(|C|~2|U|log|U|);最后用一个例子,说明算法的有效性。
3.
First-ly,the decision power used to measure the importance of condition attributes,according to which the condition attributes are sorted in anascending order.
首先使用决策强度来度量条件属性的重要性,并据此对条件属性按照属性重要性从小到大排序,然后按排序后的顺序,考察每个条件属性的所有断点,将冗余的断点去掉,从而将条件属性离散化。
补充资料：G(?)teaux梯度
G(?)teaux梯度
Gateaux gradient
‘凌如以梯度【珑加倒优脚曲斌;raTo rp”脱.TI，田-咖李回H的俘甲f夺丁卓x0牛的 H中与f在x。的C自妞.玫导数(G云姗uxderi珊tiVe)f。(x。)相等的向量.换句话说，G舀teaux梯度由公式 f(x。+h)二f(凡)+(无(x。)，h)+。(h)定义，其中。(th)/t~0，当t~0.在”维Eodid空间中C冶姗以梯度f。(x。)为具有坐标 了叮(x。)___盯(凡)、 \口x:”口x，了的向量，并简称为梯度(脚djent).C冶如ux梯度概念可以推广到下列情形:X为侧组日的n流形(有限维)或无穷维Hilbert流形，而f为X上光滑实函数.f在其C冶如以梯度方向上的增长大于过此点任何其他方向的增长. B.M.THxo栩叼Po.撰郑维行译沈永欢、王声望校
说明：补充资料仅用于学习参考，请勿用于其它任何用途。
参考词条