1

I'm a 4th year CS and Stats student with a 3.86/4.00 GPA in the US and about to graduate soon. I want to apply to top-35 ML Ph.D. program in the US. As we all know, ML Ph.D. has become insanely competitive, and especially with the economy, it gets more competitive every year.

I'm debating if I should apply in December 2024, December 2025, or both. I'm doing research with 2 different professors and they are happy to write me a letter of recommendation, but I only have two papers in prep. So, I'm not sure if my application will be strong enough for December 2024 and I should just try and have more publications and apply in December 2025. This will be a conversation that I will ask both professors, but I'd also love to hear more input.

cag51
  • 67,924
  • 25
  • 181
  • 247
dlu
  • 21
  • 2
  • For your very last sentence, this may be helpful. Your overall question may be closed as "strongly depending on individual factors", but it may attract answers to "what to keep in mind in thinking about delaying an application to a Ph.D. program". – Stephan Kolassa Mar 28 '24 at 08:14
  • Frankly, I don't see how your application will get particularly 'stronger' the longer you wait. Did you apply for admission this fall? Why or why not? – Jon Custer Mar 28 '24 at 13:35
  • @StephanKolassa Thank you so much! – dlu Mar 28 '24 at 21:28
  • @JonCuster My reasoning is to work on more publications before applying (stronger application). I didn't apply 2023 fall since I knew my application wasn't strong enough. Stephan did provide a really interesting post about delaying applications that I will read. – dlu Mar 28 '24 at 21:31
  • 2
    Your only disadvantage for every application season where you're qualified is the time investment and the application fees. You miss 100% of the shots you don't take. – user176372 Mar 29 '24 at 02:20
  • 1
    Your question had accumulated some close votes, so I suggested some edits. We are not career counselors, and future readers will not have your exact statistics, so we frown on including too many details. But the underlying question is probably “on topic” here. – cag51 Mar 29 '24 at 06:01

2 Answers2

0

I see no reason to wait and lots of reasons not to. Next year will be just as competitive, and you are probably a decent candidate having some papers in their final stages. This assumes your professor(s) will comment on their value.

However, don't restrict your applications to just the top 35. Make it a bit broader to (greatly) increase your chances. You can get a fine education at any R1 institution in the US and similar places abroad. Waiting has few upsides and you are just guessing about where you sit now and will then.

Note that delaying your career probably isn't optimal. Also, think of the issues at the other end of your doctoral study. Do you expect the job market to be better if you wait. I was late in finishing and the job market (math) completely fell apart in the interim.

Buffy
  • 363,966
  • 84
  • 956
  • 1,406
0

Yes, you should apply, and if you're worried about getting in, apply on multiple cycles. It will not take much of your time to apply in December of 2024, and it seems to me that your file is strong. I push back slightly against the "As we all know, ML Ph.D. has become insanely competitive." At my university, we send plenty of students into PhD programs for machine learning, stats, data science, and computer science. Just last year, a student graduated with a record weaker than the one you are describing, and got into a PhD program. I think the landscape is less competitive than you are afraid of. That said, it's wise to apply to a range of programs, and include some "safety schools." I encourage our students to apply for around 10 programs. Some of our students get a master's degree and then go on to a PhD, but ones with a strong record like yours are usually fine going directly into a PhD.

One last thing. Currently, someone with expertise in machine learning, even at the undergrad level, can get a good quality and high-paying job right out of undergrad. For this reason, fewer go into PhD programs, compared to fields like math. Similarly, for the student's entire time in grad school, they will be tempted to leave and go make lots of money at a company. Lastly, when they finish their PhD, they will be tempted to leave academia. For this reason, there are not enough professors with expertise in machine learning, and that's creating a bottleneck where students who want to learn this topic are unable to get into courses on it. It's essential to the success of the machine learning community that we reserve slots in PhD programs for people who want to go into a career as a professor. The same situation in CS is described in the wonderful article Conserving the Seed Corn.

If you would like to be a college professor, I'd mention that strongly in your application.

David White
  • 669
  • 2
  • 11
  • I don't want to "out" my former student so I will mention that the PhD program he got accepted to (after 4 years of undergrad) was ranked in the top 50 but not the top 30. – David White Mar 29 '24 at 11:50