Drop Maslow's Hammer or not

Abstract

There is an old saying in America (a.k.a "Maslow's Hammer") "If all you have is a hammer, then every problem you see looks like a nail." While the techniques of machine learning (ML) has attained notable success in the artificial intelligence, and is becoming trendy in the area of resource management for D2D communications, we argue, in this article, that it may not be optimal to use the ML "hammer" to hit the D2D resource management "nail". We support this argument by (1) formulating a D2D resource management problem in terms of the energy efficiency, (2) presenting a ML algorithm to tackle the formulated problem together with testing results, (3) discussing the issues observed in the ML approach, and finally (4) abandoning the ML approach and developing traditional logic-based algorithms to solve the problem. Simulation results demonstrate that the solutions provided by the traditional techniques are effective and efficient.