Is it really no free lunch?

An inquiry of the basic concept that contradicts the universality objective of learning theory. The article is imcomplete by the way

Of anything, we are typically concerned of creating a universal framework - for example, in a meeting of which we require the unification of voices and resolution; of mathematics where category theory and a lot of others try to unify mathematics under the singular lens; or in physics, there exists the quest for the theory of universal physical law, of which the standard model almost, almost claims such title in the process. The same is true for learning theory. Specifically, it is the question as to ask, “Will there ever be a universal learner \(\mathcal{L}\) that will be able to approximate everything, to a given degree of correctness?”. The villain in such story, is then, the no-free-lunch theorem. (see (Ho and Pepyne 2001), (Wolpert and Macready 1997), and (Wolpert and Macready 2005)).

Understanding no-free-lunch

No-free-lunch on itself is a troublesome notion requiring understanding of sort. Okay, well, let’s go from the beginning, shall we?

References

Ho, Yu-Chi, and D. L. Pepyne. 2001. “Simple Explanation of the No Free Lunch Theorem of Optimization.” In Proceedings of the 40th IEEE Conference on Decision and Control (Cat. No.01CH37228), 5:4409–4414 vol.5. https://doi.org/10.1109/CDC.2001.980896.

Wolpert, David H, and William G Macready. 1997. “No Free Lunch Theorems for Optimization.” IEEE Transactions on Evolutionary Computation 1 (1): 67–82.

———. 2005. “Coevolutionary Free Lunches.” IEEE Transactions on Evolutionary Computation 9 (6): 721–35.