Autumn 2024
Gained a broad foundation in modern AI through a sequence of practical modules spanning machine learning, search, planning, decision-making, and reasoning. The course began with supervised learning methods including linear regression, classification, neural networks, and generalization techniques, followed by topics in fairness (Group DRO), feature engineering, and optimization (SGD, backpropagation). Explored sequential decision-making through search algorithms (BFS, UCS, A*), dynamic programming, and constraint satisfaction problems. Studied Markov decision processes (MDPs), reinforcement learning (Monte Carlo, Q-learning, SARSA), and game-theoretic reasoning (minimax, expectimax, alpha-beta pruning). Covered probabilistic models such as Bayesian networks and Markov networks, focusing on inference, sampling, and learning with the EM algorithm. The course concluded with a unit on logic, including propositional and first-order logic, inference rules, and automated reasoning techniques.