AI in Networks (ICE4420P)
Course Curriculum
This course aims to motivate the students towards the applications of AI in 5G and beyond networks. The course will enable the students to learn the foundational mathematics and algorithms, to be applied in wireless networks. The following modules will be covered in this course:
Evolution of networks: Introduction to data, data-driven intelligence; evolution of networks: Introduction to networks: circuit switching, packet switching, TCP/ IP, VOIP, network layers; need for AI in Networks; Practical applications of AI at different network layers; Understanding cross layer network design (concepts of user quality of Service, routing, scheduling).
Statistical learning: Basics of probability and random variables, understanding datasets, KL-divergence minimization and empirical loss minimization, Parametric modeling, Bayes classifier, ML-MAP estimator, bias-variance tradeoffs, linear models; Non-parametric modeling; Optimization and need for optimization in machine learning.
Convex optimization: Concepts of covex sets, functions, L-smoothness, μ strongly convex functions; Hierarchy of convex optimizations: Linear programming, quadratic programming, second order conic programming, semi-definite programming; Lagrangian function, properties of Lagrangian, Dual function, and dual problem formulation, Weak duality, Strong duality; Slaters condition for guaranteeing strong duality, KKT conditions; Gradient descent, stochastic gradient descent algorithms; 2nd order descent methods
Non-convex optimization algorithms: Approximate gradient and hessian algorithm, Successive approximation methods (linear and quadratic approx.), Coordinate descent methods (block coordinate descent and block successive upper bound minimization), Branch and bound method; Heuristic algorithms: Simulated annealing, Genetic algorithms, and Ant Colony Optimization
Use cases of AI in wireless networks: Basics of wireless networks, introduction to the wireless channel, AWGN noise, fading noise; Applications of AI: Wireless channel estimation and prediction in 5G+ networks, Research studies on Load-Energy Balancing in Wireless networks, Scheduling and Routing in Wireless networks, Receiver design and understanding SNR and bit error rate in BPSK, QPSK systems.
Use cases of AI in object estimation, tracking, localization: Parameter estimation using discrete time Kalman filter, practical examples on object tracking, object localization in wireless networks.
Note: Programming assignments/ hands on project will be given for modules 2-6.
Grading
Assignments/ Quiz: 40%
Programming exercises: 30%
Final exam: 30%
References
[1] Kurose, J., & Ross, K. (2010). Computer networks: A top down approach featuring the internet.
[2] S. Boyd and L. Vandenberghe, Convex optimization. Cambridge university press, 2004.
[3] A. Antoniou and W.-S. Lu, Practical optimization: algorithms and engineering applications. Springer, 2007.
[4] CVX: A quick start (MATLAB) https://cvxr.com/cvx/doc/quickstart.html;
Python: https://cvxopt.org/userguide/solvers.html
[5] S. Boyd, Subgradient methods https://web.stanford.edu/class/ee364b/lectures/subgrad_method_notes.pdf
[6] Hong, Mingyi, Xiangfeng Wang, Meisam Razaviyayn, and Zhi-Quan Luo. ”Iteration complexity analysis of block coordinate descent methods.” Mathematical Programming 163, no. 1 (2017): 85-114. https://arxiv.org/abs/1310.6957
[7] Boyd, Stephen, and Jacob Mattingley. ”Branch and bound methods.” Notes for EE364b, Stanford University 2006 (2007): 07. https://see.stanford.edu/materials/lsocoee364b/17-bb_notes.pdf, https://see.stanford.edu/materials/lsocoee364b/17-bb_slides.pdf