How do you recognize a promising machine learning Phd student? originally appeared on Quora - the knowledge sharing network where compelling questions are answered by people with unique insights.
Being smart helps, but that's only part of the story. That said, there are a few things that make a good student.
- Creativity and good taste in research This is really important. Being able to identify important problems and inventing new stuff is what makes a researcher great. You can learn this, to some extent. There are techniques, e.g. to abstract away from a basic idea, to try to discover general patterns, to ask yourself whether solutions make a difference. But this also has to do with the ability to take a step back and ask weird and new questions.
- Good time management Lots of students have great ideas. And then they pursue 10 of them. Simultaneously. This usually fails. It's a bit like with startups. If they don't have focus or pivot too many times they fail. One student I had had a lot of things going on and nothing finished. Then his wife got pregnant. This gave him focus and he did an outstanding job and got lots of papers out (and won prizes). In short, do a project. Finish it. Write the paper. Then delve deeper or move on to the next one.
- Independence without being stubborn This one is tricky. Students without an opinion don't get very far since they usually also won't have great ideas. And it isn't fun to discuss stuff with someone who agrees to everything. On the other hand, if they shoot all ideas down, it doesn't help either. The balance probably depends on professor and student.
- Being proactive Professors have zero available time. So, the only way to get on my schedule is to push. If I have to pull, it probably won't work and you'll fall through the cracks at some point.
- Excitement This is your PhD / Master's degree after all. If you aren't excited, talk to your professor and make sure this gets fixed. Everyone's happier after that.
Last, this entire thing is a bit like dating. There isn't a universally best student. Or a universally best professor. This is good since otherwise the best students would be hounded down by 100s of professors (and vice versa). Sometimes things work perfectly, sometimes they don't. And you can always 'break up' and find another adviser. I've seen this happen with great effect in both directions.
- Computer Programming: What is the Parameter Server?
- Machine Learning: Are stochastic variational approaches the way to do large scale Bayesian ML or do you see any hope of scaling up MCMC-based algorithms?
- Academia: Would you encourage a MS graduate in industry to get back to academia to pursue a PhD at 30?