lingduoduo / Reinforcement-Learning

Fundamentals

• Agent
• Enviornment
• State
• Action
• Reward
• State Transition
• Return
• Value Function
• Action Value Function
• State Value Function

Suppose we have a good policy f(a|s)

• Upon observe the state s_t, ramdom sampling : a_t ~ f(.|s_t)

Suppose we know the optimal action-value fucntion Q(s, a)

• Upone observe the state s_t, choose the action that maximizes the value: a_t = argmax_a Q(s_t, a)

https://github.com/DeepRLChinese/DeepRL-Chinese

conda create -n  reinforcement-learning python=3.10
pip install -r requirement.txt
conda install -c conda-forge jupyter notebook