ExplainabilityTransparencyTrustworthinesscs.NI
Prioritized Value-Decomposition Network for Explainable AI-Enabled Network Slicing
Shavbo Salehi, Pedro Enrique Iturria-Rivera, Medhat Elsayed, Majid Bavand, Raimundas Gaigalas, Yigit Ozcan, Melike Erol-Kantarci
January 27, 2025
ICC 2025 - IEEE International Conference on Communications
Key Takeaways
-
Explainability improves trust in AI-driven network management: The paper demonstrates that breaking down AI decision-making into understandable components—showing how individual resource allocation decisions affect network performance—helps operators understand and trust automated systems managing critical infrastructure.
-
Decomposing complex decisions into simpler parts yields better results: The Prioritized Value-Decomposition Network improved throughput by 67% compared to independent methods and reduced latency by 35%, suggesting that making AI decisions more transparent doesn't sacrifice performance—it can actually enhance it.
-
XAI addresses a real gap in machine learning for critical systems: The authors identify that while ML algorithms show promise for managing network resources, their "opacity in decision-making" has been a barrier to adoption, highlighting why explainability matters for responsible deployment of AI in infrastructure.
-
Responsible AI enables better human oversight: By providing insights into how the system allocates resources, PVDN allows network operators to "better understand, trust, and optimize" management strategies, demonstrating that explainable AI supports human decision-makers rather than replacing their judgment.
Abstract
Network slicing aims to enhance flexibility and efficiency in next-generation wireless networks by allocating the right resources to meet the diverse requirements of various applications. Managing these slices with machine learning (ML) algorithms has emerged as a promising approach however explainability has been a challenge. To this end, several Explainable Artificial Intelligence (XAI) frameworks have been proposed to address the opacity in decision-making in many ML methods. In this paper, we propose a Prioritized Value-Decomposition Network (PVDN) as an XAI-driven approach for resource allocation in a multi-agent network slicing system. The PVDN method decomposes the global value function into individual contributions and prioritizes slice outputs, providing an explanation of how resource allocation decisions impact system performance. By incorporating XAI, PVDN offers valuable insights into the decision-making process, enabling network operators to better understand, trust, and optimize slice management strategies. Through simulations, we demonstrate the effectiveness of the PVDN approach with improving the throughput by 67% and 16%, while reducing latency by 35% and 22%, compared to independent and VDN-based resource allocation methods.