强化学习原理二 BasicConcepts
- 状态，State
- 状态空间，State Space
- 行动，Action
- 状态转换，state transition
- 策略，Policy
- 用数组或者矩阵表示这样一个策略
- 奖励，Reward
- 不确定的话，表格就不适用了。这个时候就要用数学来表示：p(r=-1｜s1,a1)=1 and p(r!=-1|s1,a1)=0
- 轨迹，Trajectory A Trajectory is a state-action-reward chain.
- 返回，return 可以用来评估一个策略好还是坏
- discount rate
- discounted return
- Episode
- terminal states
- MDP Markov Decision Process
- Sets
- State
- Action
- Reward
- Policy
- Probablity distribute
- Markov property
- Sets