Course details
Stochastic Optimization and Reinforcement Learning
MATH 60623A
This course presents a unified framework for decision-making over a planning horizon in an uncertain environment, focusing on stochastic optimization techniques. Several classes of such problems are studied, together with the corresponding solution methods. Each of the classes of problems is motivated and illustrated using real applications of business analytics.
The course equips students with a diverse toolkit for tackling sequential decision-making problems under uncertainty using stochastic optimization. Students learn how to model these problems effectively and select the most appropriate solution strategies, ranging from classical optimization methods to cutting-edge techniques leveraging neural networks and reinforcement learning. The course requires a strong understanding of algorithmic thinking and computer programming.
Themes covered
Derivative-based / derivative-free stochastic optimization
Markov decision processes and dynamic programming
Approximate dynamic programming
Reinforcement learning
Neural networks and deep reinforcement learning
Two-stage / Multistage stochastic programming
Policy design for sequential decision-making problems
Important notes
Course in French : MATH 60623
Course code
MATH 60623A
Subject
Mathematics
Program
Master of Science (MSc)
Location
Côte-des-Neiges
Instruction mode
On-site learning
Credits
3