Learning algorithms from circuit lower bounds
This work provides a new characterization of learning algorithms for theoretical computer scientists and extends the natural proofs barrier, which is an incremental theoretical advance.
This paper explores the relationship between efficient learning algorithms and constructive circuit lower bounds. It establishes that if errors in many p-size circuits for hard problems can be found interactively, then p-size circuits can be PAC learned over the uniform distribution with membership queries by subexponential-size circuits, and vice versa.
We revisit known constructions of efficient learning algorithms from various notions of constructive circuit lower bounds such as distinguishers breaking pseudorandom generators or efficient witnessing algorithms which find errors of small circuits attempting to compute hard functions. As our main result we prove that if it is possible to find efficiently, in a particular interactive way, errors of many p-size circuits attempting to solve hard problems, then p-size circuits can be PAC learned over the uniform distribution with membership queries by circuits of subexponential size. The opposite implication holds as well. This provides a new characterisation of learning algorithms and extends the natural proofs barrier of Razborov and Rudich. The proof is based on a method of exploiting Nisan-Wigderson generators introduced by Krajíček (2010) and used to analyze complexity of circuit lower bounds in bounded arithmetic. An interesting consequence of known constructions of learning algorithms from circuit lower bounds is a learning speedup of Oliveira and Santhanam (2016). We present an alternative proof of this phenomenon and discuss its potential to advance the program of hardness magnification.