Isn't gbm just trees? Been using it in spark for close to an year and never bothered to learn the difference. I just run both and choose the best algorithm.
I have been working on ML for close to two years and I can't answer most of these questions. Guess I am not cut out for Google
If you apply for a ML SWE role, these are unlikely to be the questions you get, though with a company as large as Google, who knows what your interviewer will bust out.
GBM is trees, and I think the question is to explain why you would want trees, but I think the trick for fitting arbitrary differentiable loss functions in a GBM is pretty nifty.
I was the technical recruiter for IBM and I always asked abstract technical questions rather than something that can be googled