Discrete Optimization

Discrete Optimization

Institution: University of Melbourne (Coursera)

Date: March 4, 2014 - May 6, 2014

Bio: This course takes a unique approach in that it makes all of the course content available upfront. Students are given a set of assignments including the Knapsack problem, Graph Coloring, Traveling Salesman, Facility Location, and Vehicle Routing; each of which has a set of problems ranging from dozens of items to thousands (ex: a 33000 city TSP problem). They are also given a series of lectures on Constraint programming, Local Search, and Linear & Integer Programming; then have to figure out how to apply the lecture material to solve the problems.

Unfortunately, this course seems to focus more on methods as appose to implementation. The lectures focus on tweaks and improvements to heuristics but never instructs the student on best practices or even basic structure of the main programming topics those improvements are built on top of. The worst case, linear and mixed integer programming, is addressed entirely from a mathematical perspective with only a few vague references to matrices being involved in the programming implementation.

The forum consensus seemed to be that if you actually wanted to learn linear programming, the best external material to help you would be to take a course that actually teaches linear programming; or simply implement a commercial or open source linear programming solver.

Accomplishments: I completed all of the problem sets for all of the assignments, achieving optimal or 'good enough' (usually 1-2% within optimality) for most all of the problems. This scored me 85% of the total maximum possible points in the course.

Leave a Reply

Your email address will not be published. Required fields are marked *