QuestionsQuestions I am occasionally asked:
- Q: Can you help me get accepted into the MSc/PhD program?
A: Probably not; your best bet is to apply through the standard channels.
- Q: I've been accepted into the MSc/PhD program! Can you be my academic adviser?
A: Maybe, but take (current year's version of) this course first.
- Q: Can you write me a letter of recommendation?
A: If I am personally familiar with your work, then yes. Otherwise – if it's based on coursework alone – it will be very generic and probably not worth the effort, for either of us. All of my letters are confidential, and under no circumstances will I agree to release a plaintext recommendation directly to the recommendee.
Questions I have asked, answers welcome:
- ☑Is there a bound on the VC-dimension of disambiguated classes?
- ★Is there a useful notion of "average margin" for SVM?
- ★What are the statistical benefits of dimensionality reduction?
- ★Do features always induce a metric?
- What are sample lower bounds for SRM?
- Are sums extremal for subgaussian concentration?
- How does VC dimension behave under projection?
- Is there a set of Lebesgue density bounded away from 0 and 1?
- How many DFAs accept two given strings?
- Is there a "truly" PAC learnable family of formal languages?
★ indicates a current research interest