Hacker News Reader

One of the authors of this paper, Thomas Vidick, taught the Advanced Algorithms class I took at Caltech a few years ago. The dude is a genius, but the class kicked my ass. He would basically assign recently solved problems from the theory community as homework.

View on HN

Yeah, assigning open/recently solved problems seems to be relatively common in graduate theory/aglo courses, at least relative to other subfields of CS and math. Not sure why. It always struck me as not the most productive use of problem set time. But then, what do I know about teaching theory :)

These classes often encourage you to collaborate - with others in the class, and also with outside resources. That is: they would encourage you to go read the paper, so long as you cited it as a source and provided a proof in your own words you had reconstructed from memory. The idea is to encourage engagement in the community and teach people to build off of existing results, a skill often overlooked in undergrad where that would have been considered cheating.

I took a class in undergrad that was similar. I think the average score on the final was 35 / 100. Of course the grades were on a curve and nobody got less than a B as long as you made an honest effort on the homework, but that class was definitely a blow to the ego for a lot of smart kids (myself included).

Reply on news.ycombinator.com