Machine Learning Over Wireless Edge Networks

Catalog description

This course aims to introduce the students to the paradigm of “intelligent wireless edge network.” The course will cover: (i) Machine learning by wireless networks and (ii) Machine learning for wireless networks.  In the former topic, the students first become familiar with some basic concepts in machine learning (e.g., convex/non-convex loss functions, smoothness and strong convexity properties, stochastic gradient descent, and convergence analysis). Afterward, distributed machine learning techniques, such as federated learning, will be covered. The students will then become familiar with server-based and serverless distributed machine learning techniques and the impact of edge device heterogeneity (e.g., in terms of communication, computation, and proximity) on the performance of distributed machine learning methods. In the latter topic, the course will focus on the applications of machine learning methods (mostly reinforcement learning) to improve the efficiency of emerging wireless networks. In particular, learning-based network optimization for edge computing, drone/UAV-assisted systems, and age-of-information minimization will be covered.