I am familiar with planning. Given a description of the possible initial states of the world, a description of the desired goals, and a description of a set of possible actions, the planning problem is to synthesize a plan that is guaranteed to generate a state which contains the desired goals. The problem can be formalized as a MDP, the known transition and reward functions allow us to determine the next state and the next reward, according to the current state and the action to execute. But, I do not really know model-based machine learning. And, when I read articles about this subject, I do not understand the difference with planning. Thank you.

  • $\begingroup$ In RL (which seems to be what you’re asking about) you would use a model to do the planning, ie rollout environment steps without actually interacting with it. $\endgroup$
    – David
    Aug 24, 2022 at 13:33
  • $\begingroup$ Thank you for your response. Yes, but in planning, we also use a model. So I don't understand the difference $\endgroup$
    – laura.znnt
    Aug 24, 2022 at 14:12
  • $\begingroup$ I think this question is a bit off. If you do know planning, you should clearly be able to understand the difference to model-based machine learning -- provided you know what the latter is. So it seems you are actually simply asking "What is model-based machine learning? Please explain!". This seems to be an overly abstract question and I do not expect that anybody is going to answer it for you unless you make some effort in explaining what you think it is and why it's still unclear to you where the differences lie. (I would not be surprised about downvotes right now!) $\endgroup$
    – Prof.Chaos
    Aug 28, 2022 at 5:09


You must log in to answer this question.

Browse other questions tagged .