Thursday, April 26, 2012

Coursera's Design and Analysis of Algorithms I — Finally finished!

I just finished the final exam for this class, taught by Tim Roughgarden of Stanford. At the moment I'm bummed out because after spending six or seven intense hours studying and arranging materials for and then taking the final I earned a 24.00 out of 30.00 on what was essentially an open-Internet test with two extra hours to research. (I had almost as many correct at the 0:58 mark as I did at the end, so I should have had better return from the two remaining hours.) Of the three questions I missed, I had no clue about one and misread or made an avoidable error on the other two. Still, I count myself somewhat lucky to get 80% on the final. I accrued technical debt throughout the course, at least until the point that &dayjob; ended and coincidentally the class moved on to graphs. I did not understand a handful of aspects of analysis or related mathematics on early topics (and that's just counting what showed up on the problem sets), but thanks to multiple choice answers and two chances to submit for the problem sets (along with programming assignments which were tedious but not tricky) I had 100% going into the final and ended the class with 94%.

Putting the same amount and kind of effort into the Stanford class, I would have been lucky to get 80% overall, though in effect it would be a different person taking that class at Stanford so all bets are off. The grading aspect isn't really so interesting anyway because there is little if any reputation at stake. (Will anyone try to find a piece of paper taped to my fridge for each MOOC whose forum I show up in?)

Some hundreds of people on the class forum supported the notion that Prof. Roughgarden is a teacher of some talent, and I concur. Thus, some useful artifacts were created -- namely, lecture videos, homework assignments of different types, etc. with matching idiosyncrasies. (For the moment let's ignore the current Stanford/Coursera agreement that puts in question these very artifacts.) What I'd love to see besides the obvious fixes (minus 2 for spelling, Tim!) are layers of crowd-contributed material à la MST, Pop-Up Videos, etc. to provide crowd-vetted context, pointers to remedial materials on the specific topic or terminology, etc. I should be able to pause the lecture and read or reference other materials directly.

Is a CS Education Movie Database with precomputed Norvig Number far behind?

No comments: