Pervasive AI for IoT Applications (Tutorial lecture)

Traditional cloud-based IoT architectures suffer from many issues, including scalability, communication and computational efficiency, in addition to privacy. This motivated the need for new emerging trends such as Edge, Fog, and Pervasive Computing, where we merge hierarchical computing with efficient communication, and leveraging learning-based distributed optimization, in order to resolve many of the issues highlighted above. In this talk, I will highlight the motivation behind pervasive AI models for Internet of Things (IoT), and cyber-physical systems (CPS), in light of traditional cloud-based architectures. Then, I will discuss some contributions we have recently published regarding distributed inference/classifications in IoT, and multi-drone systems, taking into considerations privacy and mobility of network users. I will also cover recent contributions regarding distributed learning scenarios using multi-agents and federated learning architectures that address heterogeneous user data to improve the learning performance, and outcomes in distributed networks.