Cornell CS PhD vs UIUC Stat Masters??

[confusedcow]

(crosspost from Decisions for more visibility)
I'm super grateful for my opportunities, and I feel like this is a bit of a weird dilemma to have, but I would love to to hear everyone's thoughts!

Long story short (maybe not that short), I'm currently working as a software engineer, and I have come to the unfortunate realization that I'm not the biggest fan of writing code, so I'm interested in doing research. I have a CS undergrad and a pretty decent applied ML background, but I'm interested in pursuing more "theoretical work." However, I am also a confused human who needs to try something before deciding whether or not they like it (aka what I think I want and what I actually want can be very different things). Since I don't really have any experience with theory, my research direction is pretty unclear. I also barely took any math courses in undergrad so I feel a bit under-qualified to pursue a more "theorem and proof"-based research style.

Cornell CS PhD (funded)

• a great program! am I crazy for thinking about turning it down??
• potential advisor does not seem like the best fit in terms of research interests, but he does seem flexible to new ideas
• I think a few professors do really interesting work, but they did not seem super keen on taking me as a student due to my lack of theory/math bg
• I'm not sure if I actually like CS or if I'm just familiar with it? I've tried systems + applied work and felt m e h, and theory seems like the last thing to try
• so far from any cities (or airports)

UIUC Stats Masters (also funded)

• gives me opportunities to focus on learning and figuring out what I want to do
• I'm open to pivoting and focusing on statistics, but I feel like I might be throwing away my experience in CS
• the courses seem really interesting, but what if I don't like them or what if I fail out LOL
• not as "reputable"? definitely not as rare of an opportunity as Cornell PhD I think
• would require either reapplying to PhDs in 2 years or transferring into UIUC Stat PhD in one year (if they take me)
• also in a cornfield, but my SO is there!

Thanks for reading :)) I'd love to hear your initial impressions, and let me know if there's something really obvious that I'm missing l o l i've been overthinking this way too much and my brain is tired

[TheMeanGene]

Hi, since this is quite the unique situation you are considering, could you provide some more information about what is perhaps drawing you to make a switch? You mention that going to UIUC would require reapplying to PhDs in 2 years, so it seems apparent that you want a PhD, but what for precisely? Are you intending to go to academia? Aside from the "theorem and proof"-based research style, is there a more concrete area of statistics you may be interested in exploring, e.g. ML, networks, etc.? The same goes for CS, what type of research from the theory side are you intending to pursue? Overall, there is quite the contrast between the two decisions so it could help to be a little more specific.

[confusedcow]

Sure, happy to elaborate! I currently work on probabilistic inference, and I've really enjoyed learning about probabilistic models and understanding the theory behind how different inference methods work. I'm generally interested in uncertainty quantification and robustness for machine learning, although I'm open to other ideas as well. My potential Cornell advisor typically approaches scalable machine learning from an algo x systems perspective, the systems side of which I'm not currently thrilled by. There are a few professors that I'm interested in who work on decision-making under uncertainty / RL and causal inference, but they don't seem interested in taking me. I think that developing a solid stat/math background would make me more qualified to do this type of research, but I'm not sure if it would be possible for me to develop this during my PhD (while being expected to produce a high volume of research), or if I should focus on bolstering my background through a master's degree first.

[TheMeanGene]

I am not terribly familiar with how CS PhDs operate aside from the focus being more on the research output as you mentioned. My understanding is that they expect you to produce research from the get-go and learn as necessary to solve whatever problems or topics that come up in your research. I'm not sure how much opportunities you may have to pursue the type of coursework you may be interested in learning (especially if you have to take some more "irrelevant" cs courses to fulfill requirements).

Since your UIUC masters would be funded and only a 1-2 year commitment, it might not actually be all that bad of an option to give yourself an opportunity to explore a more theoretic side of things while also setting yourself up to be a more rigorous candidate for theoretically driven research. Both options clearly have their pros and cons, but I'd think that UIUC is a more "safe" option. Cornell CS definitely is prestigious, and this opportunity won't be easy to come by again perhaps (should you turn it down), but it seems like your experience primarily weighs on whether you will be able to do the more theoretic type research you desire. This to me seems like a bit more of a gamble.

Also note that I'm not taking in any external factors that you may also wish to consider such as location, stipend, SO, etc. These are definitely something you have to consider for yourself and determine what you value and are willing to give up. Obviously keep in mind that this choice is ultimately up to you and others definitely may have differing opinions as opposed to mine (especially as someone who is in the statistics field and not CS).