Abstract
The shift to AI-native 6G networks demands autonomous slicing strategies that can adapt to diverse and evolving edge and IoT service needs. Two paradigms have emerged: Learn to Slice (L2S), where AI optimizes network slicing for general services, and Slice to Learn (S2L), where slices support AI model training, often offloaded from Internet of Things (IoT) devices. Existing S2L approaches typically optimize communication or computation in isolation. This paper presents the first unified framework that jointly optimizes communication resources, computation capacity, and AI hyperparameters to maximize the average accuracy of multiple concurrent AI services. We address the complexity of this joint problem by applying L2S-inspired techniques to enhance S2L, introducing two autonomous agents: EXP3 from online convex optimization and DQN from deep reinforcement learning. Extensive experiments demonstrate and contrast the effectiveness of these agents in maximizing aggregated AI accuracy, supporting knowledge transfer, and sustaining robust performance under adversarial and long-term conditions, thereby enhancing the realization of zero-touch network management for AI services in 6G networks, supporting resource-constrained IoT. © 2025 Elsevier B.V., All rights reserved.

Abstract
The forthcoming 6G networks will embrace a new realm of AI-driven services that requires innovative network slicing strategies, namely slicing for AI, which involves the creation of customized network slices to meet Quality of Service (QoS) requirements of diverse AI services. This poses challenges due to time-varying dynamics of users' behavior and mobile networks. Thus, this paper proposes an online learning framework to determine the allocation of computational and communication resources to AI services, to optimize their accuracy as one of their unique key performance indicators (KPIs), while abiding by resources, learning latency, and cost constraints. We define a problem of optimizing the total accuracy while balancing conflicting KPIs, prove its NP-hardness, and propose an online learning framework for solving it in dynamic environments. We present a basic online solution and two variations employing a pre-learning elimination method for reducing the decision space to expedite the learning. Furthermore, we propose a biased decision space subset selection by incorporating prior knowledge to enhance the learning speed without compromising performance and present two alternatives of handling the selected subset. Our results depict the efficiency of the proposed solutions in converging to the optimal decisions, while reducing decision space and improving time complexity. Additionally, our solution outperforms State-of-the-Art techniques in adapting to diverse environmental dynamics and excels under varying levels of resource availability.

Abstract

Abstract

Abstract

Abstract