Subproblem learning: A misnomer