Slides for: Joint Optimization of Wireless and Computing: From Computation Offloading to Edge Learning in 6G

The research on 6G has kicked-off all around the world, opening the floor for real paradigm shifts. While the question of what will 6G be is still completely open. It is unanimously agreed that Artificial Intelligence (AI) and Machine Learning (ML) will be native components of such networks. This comes with two facets: i) Learning to communicate, by which AI/ML tools are exploited for flexible and autonomous network optimization, and ii) Communicating to learn, by which 6G is exploited as an efficient AI platform. Focusing on the latter, this webinar aims at providing vision, challenges, and tools that will effectively and efficiently enable ML/AI at the edge of wireless networks, with target performance in terms of energy, delay, learning/inference reliability. Starting from the concept that edge learning and inference are computation offloading services supported by Multi-access Edge Computing, a journey through the joint optimization of radio and computing resources will be presented, with target performance in terms of energy and delay. Performance trade-offs will be discussed through theoretical and simulation-based findings. Building on the acquired vision and tools, the webinar will then focus on edge inference, which requires a third axis: inference reliability (e.g., accuracy and confidence).