Machine Learning over Networks
Carlo Fischione,

January 19th 10-13, January 20th 10-12

Microsoft Team: PHD ICT – Seminars and Courses


Short description:
One of the main characteristics of the Internet of Things (IoT) technological revolution is the generation of huge quantities of data. Such wealth of data and their use in several new IoT technologies is forcefully motivating the development of data analysis methods, namely machine learning. Currently, machine learning needs big datasets and very huge computational and communication resources. However, in IoT, data sets of any size will be distributed among several nodes (people, devices, objects, or machines) that might not be able to perform the computations and to share data.

Unfortunately, existing machine learning methods are mostly intended for proprietary or high performing networks such as in data centres. They would greatly stress the public communication networks of IoT, such as 5G wireless networks and beyond. In these networks, machine learning methods will encounter new challenges in terms of computation, bandwidth, scalability, privacy, and security.  Machine learning over networks face a lack of understanding of the fundamental methods, and poor performance of their algorithms.

In this PhD course, we highlight the need of establishing a new fundamental theory for machine learning over networks. We give an overview of the state-of-the-art and some of our recently proposed developments. The syllabus will be around the following topics:

  1. Introduction to Machine Learning for the IoT
  2. Background on Machine Learning
  3. Distributed Machine Learning
  4. Wireless for Machine Learning
  5. Co-design of Machine Learning and Wireless
Machine Learning over Networks