Syllabus & policies
Course objectives, prerequisites, references, assessment policy, and weekly plan.
To be postedGraduate course · Faculty of Mathematics
Sample course pageSemester TBA
Optimization Methods for Data Science
Optimization algorithms and numerical methods for modern data-science models.
Course overview
About the course
This course studies optimization methods that form the computational core of many data-science and machine-learning models. The emphasis is on understanding algorithmic ideas, convergence behavior, implementation issues, and the trade-offs that arise when problems are large-scale, stochastic, constrained, or nonsmooth.
The page is prepared as a reusable course template. The semester, timetable, assessment scheme, and downloadable materials can be filled in once the teaching assignment is finalized.
Syllabus
Indicative topics
The final ordering and depth can be adjusted to the background of the class and the number of teaching weeks.
- Optimization models in data scienceProblem formulation, loss functions, regularization, convexity, and optimality conditions.
- First-order methodsGradient descent, step-size selection, line search, momentum, and accelerated gradient methods.
- Stochastic optimizationStochastic and mini-batch gradient methods, sampling, convergence considerations, and variance-reduction ideas.
- Composite and nonsmooth optimizationSubgradient methods, proximal operators, proximal-gradient algorithms, and sparse regularization.
- Constrained optimizationProjection methods, penalty ideas, constrained models, and numerical treatment of feasibility.
- Large-scale numerical optimizationMatrix-free computation, conjugate-gradient ideas, stopping criteria, and practical numerical issues.
- Selected data-science applicationsRegression, classification, sparse learning, and other models chosen to illustrate the optimization methods.
Student resources
Course materials
These panels are ready to host files and links after the course schedule is finalized.
Lecture notes
Slides, handwritten notes, computational notebooks, and supplementary reading.
To be postedAssignments
Problem sets, programming exercises, submission instructions, and solutions when appropriate.
To be postedCourse project
Project topics, milestones, report format, presentation guidance, and evaluation criteria.
To be postedUpdates
Announcements
Course offering not yet finalized.
The semester, meeting times, and course materials will be updated here after the teaching schedule is confirmed.