Objectives: (1) Identify critical context information for anticipatory performance optimisation; (2) Study ML techniques to predict the context information; (3) Develop anticipatory optimisation techniques; (4) Assess the impact of anticipation accuracy, and resilience to data corruption.
Expected Results: (1) Identify ML algorithms to predict the evolution of key context information (e.g., channel gain, traffic generated by a source, nodes mobility); (2) Develop optimisation techniques based on such predictions. Performance gain to be proved through simulations
Objectives: (1) Develop a scalable network-wide unsupervised learning framework based on deep networks and probabilistic generative models; (2) Understand the fundamental limits of such learning systems and the related network performance, along with the importance of the various sensory data in the optimisation of key networking functionalities; (3) Build a suite of learning tools for specific network slicing objectives.
Expected Results: (1) Design a suite of learning tools for local and global optimisation objectives; (2) Define the strategy to effectively support the learning process and the network automation by exploiting the network softwarisation paradigm.