Discrete Optimization

Discrete optimization is a quasi-self-paced programming course offered by the University of Melbourne through Coursera that is all about solving hard problems. Hard problems in the context of this course means NP-hard problems--problems with exponential worst-case running times. The course differs from most classes on Coursera and elsewhere on the web in that all the materials are available as soon as the course opens, but there is a final deadline for the programming assignments, so it is not a self-paced course in the truest sense. The entire course grade is based on 5 programming assignments: the knapsack problem, graph coloring, traveling salesman, warehouse location and vehicle routing. An average score of 7 (out of 10) on each part of each programming assignment is required to earn a certificate.

Discrete optimization opens with an introductory lecture series on the knapsack problem that lasts a couple of hours followed by three longer lecture series, covering constraint programming, local search and mixed integer programming. The lectures do not need to be viewed in any particular order. Similarly, students can work on the homework projects in any order they choose. This level of freedom is great for students who want to work ahead but it may make it difficult to complete the course if you don't plan ahead because the programming assignments can be very time consuming. The assignment skeleton and submission code is written in Python 2.7, but you can use languages if you want.

The professor, Pascal Van Hentenryck, is extremely energetic and passionate about the subject. He makes the lecture videos surprisingly fun for such a dense subject. The lecture videos themselves are well-made and the professor does a good job explaining the material, although I sometimes felt like the course was trying to cover too many different topics and it wasn't always clear how one would go about applying the methods in lecture to the assignments or using them without using some external package or solver. A little more instruction and direction in that regard would be helpful.

Discrete optimization is challenging course with great programming assignments that introduces many different tools and leaves them on the table for you to play with. The tools don't always with full instruction manuals, so you'll have to figure out many of the details yourself. You won't have time to apply every tool to every problem, but if you focus on one and budget your time well, you'll have a good shot at making it through.

I give discrete optimization 4 out of 5 stars: Very Good.

This course was excellently designed.

Students who prefer a rigid workflow and hand-holding will strongly dislike this course, as its content is quite challenging. However, students who like a flexible schedule and exploration will appreciate the effort placed into making this course catered just for that.

Professor Pascal Van Hentenryck gives very good explanations and makes use of visuals to help students understand the content. Furthermore, there is a leaderboard for the 6 problems which are tackled in this course, where students get to challenge themselves to produce excellent solutions. The videos are released all at once for students to go through the course in whatever way they want.

Also, this is one of the few courses where the professor is actively following the students. A weekly video is released describing the course situation and giving some interesting information.

Unfortunately, I was unable to complete this course due to the level of difficulty. I am planning to retake this course when I have gained a better understanding of the fundamental concepts.

I was really torn on this class. On one hand, it focuses on really cool problems. If you’ve ever wanted to know how best to handle intractable problems like the traveling salesman problem or the map coloring problem, this is your course. Prof Pascal Van Hentenryck’s explanations of the problems were top-notch and his enthusiasm was just contagious! Unfortunately, he chose to use only “pseudo-code” in lecture examples. This made deciphering his algorithms needlessly troublesome. Worse yet, his pseudo-code wasn’t similar to Javascript, Ruby, C, Java, Lisp or any other major language. Instead it was most similar to his own commercial language, Comet. Also, we were free to solve the problems in any order we chose, so long as we did one per week. Furthermore, we were allowed to return to previous assignments as we learned new techniques apply them for a better ranking. I loved that!

I appreciated the sense of humor, the goofy costumes even. But after the first unit, in my experience, the course becomes almost unworkable. The lectures are long, rambling, and often off-topic. The programming assignments are difficult, and very little can be gleaned from the lectures that directly helps in solving the problems. In the end, I made several time-consuming attempts at solving the 2nd and 3rd assignments, before giving up.

After the first unit, Professor Van Hentenryck does very little to show actual techniques for solving the given problems. The goal, it seems, is for the students to come up with their own. While I understand this to a point, for a newcomer to the field, this is an extremely inefficient way of doing things. I spent hours debugging a solution, only to find the solution wouldn't actually solve the problem.

If I were already versed in this area, I might well have enjoyed this format.