It seems that there are certain machine learning algorithms out there that are most suitable for “control tasks” and “controller designs”. I know that there’s a lot of different ways for the word “control” to be defined/interpreted, so I was wondering what it means in the context of machine learning? And also, if it means the same thing when describing something as a “control task” or a “controller”.
If you get what I mean thank you very much!
The control task refers to the rewarded control behavior that is desired to be improved. The controller refers to actor (for example the Neural network), and controller designs refers to the design of the controller. This design can focus on how the system learns (using neural networks), how it could be composed of existing systems (and how these systems will communicate), or whatever the designer deems to yield the most effective solution.
Excellent question as there are many conjugations “control” being used.
So as mentioned, they do refer to different things. The “control task” is concerned with the method of rewarding the controller for performing the correct task.