Re:Confession as an AI researcher; seeking advice
by stochastic_monk on Sunday November 12, 2017 @05:51PM
(#15683009)
I can only speak from experience in my CS PhD program, but my above recommendations are based directly on that experience.
We have an exam after the first two years and a similar process. It depends on what your advisor expects/wants. Mine has been flexible. And I focused more on efficient software engineering and applications for prior methods than on new research as I got up to speed on other matters. It made obvious a lot of ways I could improve them, just by being forced to look at and implement all of the details under the hood.
And, at least in my CS department, there is a very heavy emphasis on mathematics with the ML/AI folks. They coauthor a number of papers with the applied math department and the rest of their papers are mostly proofs. They'll usually back it up with proof-of-concept implementations, but in that regard, they're very much like researchers in applied math except that they use Python instead of MATLAB.
We have an exam after the first two years and a similar process. It depends on what your advisor expects/wants. Mine has been flexible. And I focused more on efficient software engineering and applications for prior methods than on new research as I got up to speed on other matters. It made obvious a lot of ways I could improve them, just by being forced to look at and implement all of the details under the hood.
And, at least in my CS department, there is a very heavy emphasis on mathematics with the ML/AI folks. They coauthor a number of papers with the applied math department and the rest of their papers are mostly proofs. They'll usually back it up with proof-of-concept implementations, but in that regard, they're very much like researchers in applied math except that they use Python instead of MATLAB.