SC 607: Optimization

Instructor

Prof. Ankur Kulkarni

Motivation

The choice to study optimization is a natural one, or as the course instructor mentions on the course page: ‘If we exist, we think. If we think, we make choices. If we make choices, we want to optimize.’
The course is mathematically rigorous and will cover fundamental optimization techniques which find wide applications in great depth. The professor does a great job of explaining complex concepts through their geometrical interpretation.

Prerequisites

Familiarity with linear algebra and real analysis along with the enthusiasm to pick up concepts quickly. Instructor consent might be required for UGs.

Course Content and Structure

(a) Review of Real Analysis
(b) Optimization in Euclidean spaces: Linear programming, Constrained Optimization, KKT conditions
(c) Algorithms and Convergence: Line search method, simplex method, Interior point method etc.

Students were divided into groups (formed by them) with each group given the responsibility of taking up one lecture using a presentation each after watching the CDEEP videos of the course. The professor would regularly help during the class, help in resolving doubts, both of the presenters and the student audience. A guest lecture from an industry expert on Optimization was taken towards the end.

Weightage

Lecture Presentation: 20%
Homework/assignments: 40%
Final exam: 40%

Feedback on lectures, tutorials and exams

(a) Lectures: Since the in-class lectures were taken up by students, some lectures were unstructured. However in such cases the professor would quickly take over and explain. The CDEEP lectures are very well taken and the professor does a great job of explaining highly abstract topics intuitively.

(b) Tutorials: No tutorials were taken as such. However the assignments (which were objective questions) were very instructive, with each problem covering an important section. Ample time was provided to solve these.

(c) Exams: Only 1 end-sem exam was taken. It was an objective paper with a time span of 3 days and no proctoring. Solving the assignments regularly should be enough to do well.

Attendance

No attendance policy was maintained. However, attending lectures regularly is advised since the course picks pace very quickly.

Difficulty level

5

Grading Stats

This is definitely a tough course with harsh grading cut-offs. Only 1 AA was allotted. Students are advised to take this course only if they’re interested in the course content.

References

S. Boyd and L. Vandenberghe, Convex Optimization, Cambridge University Press, 2004

We thank Shreyam Mishra for this review.