Comment by YeGoblynQueenne

It depends on how you mean "converge". It would certainly tend towards the full ruleset, learning more of them, or more accurate versions of them, the more examples it saw. It would not really terminate though, at least according to the learnability results from inductive inference.