Interpreting AI for Networking: Where We Are and Where We Are Going

Abstract

In recent years, artificial intelligence (AI) techniques have been increasingly adopted to tackle networking problems. Although AI algorithms can deliver high-quality solutions, most of them are inherently intricate and erratic for human cognition. This lack of interpretability tremendously hinders the commercial success of AI-based solutions in practice. To cope with this challenge, networking researchers are starting to explore explainable AI (XAI) techniques to make AI models interpretable, manageable, and trustworthy. In this article, we overview the application of AI in networking and discuss the necessity for interpretability. Next, we review the current research on interpreting AI-based networking solutions and systems. At last, we envision future challenges and directions. The ultimate goal of this article is to present a general guideline for AI and networking practitioners and motivate the continuous advancement of AI-based solutions in modern communication networks.