I'm working on a scenario/environment where I have a simulation that provides an arrangement or results of the simulation that has data in a format of samples in vectors(x,y,z,N). Let's say it maps 100 samples in a specific space based on the vectors. What I would like to do is to apply RL to find out the best possible combination of samples based on their vectors/coordinates.

How would I go about applying or mapping this simulation to find out an ideal arrangement (combination or order of samples) of the vector to achieve an optimal score?

  • 2
    $\begingroup$ You can apply RL in many scenarios, where it is not necessarily the best choice. Looks like this may apply here - you could use RL, but only as a search mechanism for a combinatiorial optimisation problem. What I am not seeing in your description, is any concept of time step or moving through a state space that is meaningful. I.e. what you care about is the combination of vectors at the end? It may help to give more details about how the vectores are used - e.g. must they be used or rejected one by one when they are presented? Also, whether there is any concept of an interim score? $\endgroup$ – Neil Slater Nov 18 '19 at 14:35
  • $\begingroup$ Hi @NeilSlater, thanks for the help! That might be the issue is that moving through a state space isn't meaningful, essentially the coordinates of the samples each time the simulation is run can produce an ideal 'build' or positioning of the samples. within a space - let's say a cube. The vectors wouldn't have to be rejected, but they would have to all be used to produce the optimal arrangement. As for the score, it's also a concept I'm working on trying to define as ideally the arrangement of the samples would produce the ideal score or best score. Does that make sense? $\endgroup$ – RandomGuest Nov 18 '19 at 15:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.