Ties Petersen.

We won twice by NOT finding the perfect solution.

Solving real-world scheduling and resource allocation problems through advanced heuristic algorithms during a competitive challenge.

( 1 min read )

A great way to start my third year at TU/e, by winning both "real-world" optimization challenges that were part of the Heuristic Algorithms course that I took last quartile, together with my teammate Henriks Meija.

It was a great hands-on experience applying what we learned to complex "real-world" puzzles. The goal of these (fictional) challenges was to find the best possible solution to the following problems:

Problem 1: Assigning bachelor students to group projects, trying to balance their teammate preferences, skills, project topics and mentor availability.

Problem 2: Designing a cost-effective plan of new Master's study tracks and marketing strategies to attract a target number of students.

We solved them by designing custom algorithms based on the heuristics taught during the course. Very simply put: heuristic algorithms don't try to find the perfect solution (which is often computationally impossible), but search through the different options efficiently to find a very good solution instead.

I've been getting more and more interested in algorithm design, and this course was a great confirmation. It's interesting to see how these techniques can be applied to large-scale scheduling and resource allocation challenges that universities and companies encounter in real life.

* Curious about the technical details? Our main approach for both challenges involved using a highly-optimized simulated annealing algorithm, which consisted of caching systems, multiple neighborhoods, and a reheating strategy to escape local optima. We also experimented with other metaheuristics, such as genetic/evolutionary algorithms and exact solvers (ILP). Want to know more? Send me a message!

×

×

Scoreboard of both challenges, where we secured 1st place in each as team O(n^30).