Action space v decision class?

[indigorose1]

What exactly is the difference between action space and the decision rule class? Do the decision rules exist in the action space or is that just how they're formed?

[jtmatterer]

A decision rule is a function that takes the data and returns an action that belongs to the action space.

[indigorose1]

Thanks!

[mblohr]

I like to think of sample space as the input to an algorithm (the decision), the action space as the output of the algorithm, and the loss function as how accurate the algorithm works.