I think many ML isn’t even engineering. Speeding up the implementation of an algorithm is, but the algorithms themselves mostly are of the “something like this seemed to work for problem P, so for problem Q, I tried this variation” type.
Yes, there may be solid math behind it that says “if your problem is of type T, this algorithm will get within Foo of the optimal solution in O(n log n) time”, but the problem is that nobody can tell whether a given real-world problem is of type T. Yet, people happily run the algorithm and if it works, it works.
Yes, there may be solid math behind it that says “if your problem is of type T, this algorithm will get within Foo of the optimal solution in O(n log n) time”, but the problem is that nobody can tell whether a given real-world problem is of type T. Yet, people happily run the algorithm and if it works, it works.