How to save state in reinforcement learning

I wanna know the way to save information of agent's state in reinforcement learning.

Answer 1

[Harsh]

Hi Ryunosuke,

To save an agent while training you can use “SaveAgentCriteria”, “SaveAgentValue” and “SaveAgentDirectory” properties of “rlTrainingOptions” object to control when and where to save an agent while training.

“SaveAgentCriteria” is used to specify the condition for saving agents during training. “SaveAgentValue” is used to specify the critical value of condition for saving agents. “SaveAgentDirectory” is used to specify the directory for saved agents.

Please refer to the following documentation to understand more regarding “rlTrainingOptions” object and its properties - https://www.mathworks.com/help/reinforcement-learning/ref/rl.option.rltrainingoptions.html?s_tid=doc_ta

After training your agent, you can save it using the “save” function. Please refer to the following MATLAB answer for more information - https://www.mathworks.com/matlabcentral/answers/816585-saving-trained-rl-agent-after-training