"What's your favourite machine learning algorithm?"
According to my friend Kyle Becker, this interview question seems to throw off quite a few candidates. Honestly, it threw me off as well: I don't think I have a "favourite machine learning algorithm", I typically feel that the best solution is largely dictated by the problem to be solved. However, when thinking a little more about this question, I must admit that there are some ideas that I find very elegant and that certainly had a big impact on the way I think about data.