[PartiallyTyped]

I wasn't thinking of DeepRL, but more on the more classical side of things with approximators other than neural NNs; but what you describe makes sense.

[bo1024]

On that side, reinforcement learning bleeds over into control theory, so you're partly right.