当前位置：网站首页>强化学习中的符号命名分析
强化学习中的符号命名分析
2022-07-16 06:19:00 【wzg2016】
Q: 代表未来收益的累计值，物理意义是: cumulate reward after seeing state S and taking action a. Q与cumulate的首发音相同。
π：通常用来代表policy，因为π与policy的首发音相似。
价值函数V与Q函数的关系：
- 价值函数仅与S相关，
- Q函数同时与S与action相关。
- 二者之间有转换关系。考虑到所有动作后的Q价值的累加和是V。
边栏推荐
- I "bring salt" for tdengine! "High price" to recruit photographed developers
- 05 gulimall VirtualBox set fixed IP for virtual machine
- 【sdx62】SBL阶段读取GPIO的状态操作
- 家族树问题
- Background running program method
- 首届京东科技合作伙伴大会召开，博云携手京东科技共创产业数字化新增长
- AcWing 3652. Maximum continuous subsequence dynamic programming
- xsslabs通关
- ViewGroup event distribution sorting
- Configure maskrcnn environment roast (geforce mx250+win10+tensorflow1.5.0 GPU version)
猜你喜欢
![[every Monday shift] - Issue 3: web development security precautions](/img/2e/64e2f7aca24abd6b68d844e0b78a3e.jpg)
[every Monday shift] - Issue 3: web development security precautions
为什么电商平台要重点关注非法经营罪，这与二清又有什么关联？
Ziguang Tongchuang FPGA development jump pit Guide (V) -- DDR3 controller IP simulation
Wechat selection and voting of applet completion works applet graduation project (8) graduation project thesis template
JVM内存模型
Eureka read-write lock fantasy, too top!
Xsslabs customs clearance
Variables in shell scripts
About mock third-party calls
出海已成大势，技术如何赋能？| ArchSummit
随机推荐
风控人不能不知的黑产大揭秘
Stonedb announces open source, why is the integrated real-time HTAP architecture the current best solution
sudo 找不到命令 command not found 解决方案
The MySQL data of the government network server has opened the network strategy with dataworks, but the test still refers to the connectivity of the database test
Mitmproxy getting started installation
Remove the problem of repeated letters (the minimum sequence of different characters)
Total sequencing problem
Failed to determine a suitable driver class problem solution
A series of questions about candy
CUDA and cudnn installation tutorial (super detailed) Uninstalling CUDA, installing nsight visual studio edition of CUDA fails, and there is no CUDA option in vs2019+cuda11.1 new project
HMM ＆ MEMM ＆ CRF
小程序毕设作品之微信评选投票小程序毕业设计（8）毕业设计论文模板
为什么电商平台要重点关注非法经营罪，这与二清又有什么关联？
【C语言】printf格式化输出及修饰符总结
德银天下港交所上市：市值39亿港元 陕汽集团是大股东
AcWing 3619. Date date processing
配置MaskRCNN环境吐槽（GeForce MX250+win10+tensorflow1.5.0 GPU版）
Function stack frame (worth collecting)
求数据流中的中位数问题
首届京东科技合作伙伴大会召开，博云携手京东科技共创产业数字化新增长