How to Apply to Graduate School

Today is March 25th, 2023. I have almost ended my graduate application. As a student who has made tons of mistake. I would like to share something with you to hopefully prevent you from making the same mistake

Profile

UG: SJTU JI ECE 3.4 + UMich CSE 4.0

Research: UMich, 1x Sys + 1x ML, no pub

Intern: None

Letter: 1x citation 20k+ US PI research letter, 1x citation 10k+ US PI research letter, 1x China teaching faculty TA letter, all strong

Application

I try to apply very aggressively, as I have plan to gap if failed to get in schools I desire.

PhD

Result

Note: Research-based master programs (e.g. CMU MLT/MSR, UIUC MSCS, UWisc MSCS) are classified as PhD programs in this table because they have similar admission process. Some explanations about thesis-based master programs:

• CMU MLT/MSR is a two-year thesis program where you pair with an advisor and do research while taking only 7 courses across 4 semesters. Depending on the funding availability, faculty members have the freedom to decide whether you are fully funded, get tuition waiver or a small stipend. Applicants are selected based on their research experience, recommendation letters and GPA without an interview. Selected applicants are required to pair with an advisor by the end of October in their first semester.
• UIUC MSCS is a two-year thesis program where you are required to start working on your thesis in your third semester. You are required to take at most 9 courses across at most 5 semesters. This program is very flexible and you may waive tuition through TA/RA. In most cases, applicants are selected by individual faculty members for research positions. However, there do exist applicants who are selected by the admission committee directly.
• UWisc consider MSCS applicants in the same pool as the PhD applicants. They do not interview. Admission is extremely competitive, where less than 50 people are chosen from 1900 applicants. Almost everyone get a PhD-level fully funded assistantship.

Q/A

Q: What is the most important thing in PhD application?

A: The most important thing in PhD application is papers. Without papers, you should not even bother applying to a good CS PhD program, unless your recommender is well-known and classify you as “the best student”. Letters of recommendation can at-most get you in an interview. If the PI interviewer is not interested in what you did, your application will still result in a failure. Also, students vary in different groups: a best student in a normal university might be a normal student in a best university. However, paper is the easiest and the most fair way of judging a student. Imagine if you have no paper, how could the interviewer even judge you?

Q: How could I prepare for a PhD application?

A: You should start very early and work hard on at most two research project in parallel.

• Start early is very important. From my observation, students admitted to a top PhD program (e.g. top 4) usually have at least 1 first author top conference paper and 2-3 years of research experience. It takes up to a year for a student to know what is research and how to do research. Within this period, you will have no paper. The best time to start research is the second semester of your sophomore year: you will have 1 year to know what is research and what you need to learn and you can publish papers on your 3rd year. By the end of your undergraduate year, you can expect yourself to have ~3 papers, with at least 1 first author paper.
• Do not push too hard on yourself. From my own observation, working on 3 research at the same time (and get 3 very strong letters) with several heavy courses result in a much worse outcome than working on 1 research without any heavy course. It is not the number of researches that matters, but the quality of your research.
• Do not try to optimize your GPA too hard. A 4.0/4.0 GPA will not help your PhD admission, but will be very harmful if you spend too much time working on your course. However, I encourage you to optimize your GPA and take hard courses before you find a research, because you should not only persuade your potential PI that you are a good student (a high GPA can be a strong predictor) but also you are a good match (courses related to his research field). DO NOT spend too much time on your course after you enter the research group, but focus on your research and try to carry out a high-quality publication.
• Find a good advisor is very important. In most cases, you will work with a PhD student, and have meetings with your PI. I encourage you to choose a PI graduated from/working at a top university (such as CMU/UCB), such PI can give you much better connection to the most influential people in your field, and greatly broaden your horizon. I also encourage you to find a senior PhD student, instead of a new PhD student who is also trying to know how to do research. The first research experience will almost certainly affect whether or not you like research, and if you end up not liking research, it will in most cases not be your fault.
• If you are rejected by some school/some PI, it might not be your fault. PhD admission is a mystery. Sometimes a PI does not have funding or become submarine, but he does not clearly state this in his personal webpage. I once emailed a PI about PhD position, but he told me that his position was filled by an undergraduate student working with him for a long time. However, if you are rejected by the majority of schools in similar levels, it might be because you are not qualified for the position.
• If your POI clearly states that he encourage you to cold-email him, then be brave and email him. He will almost for-sure reply to your email. Such email can greatly improve your chance of admission. Otherwise, there is no need to email him.
• Your research recommender can give you the most accurate feedback on your application, because he is also an expert in recruiting PhD students. My advisor told me to broaden my pool with a few easier schools, which is a nice indicator that my PhD application will result in a failure if I do not follow his advise.

Master

Result

Note: Research-based master programs (e.g. CMU MLT/MSR, UIUC MSCS, UWisc MSCS) are classified as PhD programs and are NOT included in this table because their admission process is similar to PhD programs. The following programs are course-based non-thesis professional master programs, though students can still do research and get fully funded (Stanford/UMich/UT Austin), tuition waiver or a small stipend (CMU) by working as a TA/RA.

Q/A

Q: What is the most important thing in master application?

A: Your GPA is without doubt the most important thing. Unlike PhD, you should try hard from the very beginning to optimize your GPA, no matter whether the course will be helpful to your career. I know a lot of students, like me, do not take seriously on humanity courses and political brain-washing courses. Unfortunately, such courses may prevent you from admission to top master programs. There are thousands of applicants to a single master program, and they do not have time to view your transcript one-by-one. They may throw away those who have a low GPA. Even if they do not throw away your application, they will have enough students who did well in all classes - including courses you do not like, then why do they even bother to admit you?

If you do not want to take seriously on courses you have no desire to take but you have to take for your degree, I encourage you to take it in your last year and it will not appear on your transcript when you submit your application.

However, exceptions do exist: CMU INI will waitlist too-strong applicants who clearly treat their program as a backup, and they will look into your transcript to look for relevant courses. UIUC MSCS is a research program. Admission is almost decided by individual faculty.

Q: How much do I have to pay for my master degree?

A: It depends on your school. Some universities offer full tuition waiver for students with TA/RA, such as Stanford, UMich, UIUC, UWisc, UT Austin, UCSD. Some only offer a small compensation which can fulfill your daily cost, such as all master programs of CMU.

Misc

Q: What is the most important thing in undergraduate life?

A: Your future. Other things, such as romantic relationship, should not be a factor to affect your choice. Why? You may meet much better person when you go higher. However, if you give up going higher for your current lover and you broke up, you lose everything. Club and student organization activities are almost useless for your career. If you feel unhappy, you should drop out as soon as possible. No one cares about such experience.

Q: How should I spend my undergraduate life?

A: For SJTU-JI students:

• Year 1: Take as many courses as you can. Here I mean math & CS courses online, not whatever JI offers you. SJTU, including JI, is unable to offer a diverse and high-quality CS courses and you should not rely on the almost-useless academic plan SJTU offers. You should study those US courses on bilibili/YouTube. Go through the path like this: basic programming -> data structure, algorithms, computer architecture -> operating systems, distributed systems, machine learning, parallel programming -> whatever you are interested in, including database, security, theory, etc. 1.5 years is totally enough for you. You must do assignments. As for how to treat JI courses, please try to optimize your GPA, while spending as less time as possible on chemistry and political courses. Your GPA is your entry to your 1st research, but you should spend less time on useless stuff. Better leave chemistry, humanities and political courses to a later semester, better to your last semester, to prevent them from affecting you application.
• Year 2: Find a good research advisor at SJTU. It should be someone who graduate from a good US university, such as CMU, and publish papers at 1st tier conference. On ML&CV area, Siheng Chen is definitely a great choice: he graduated from CMU, and he is surely a very strong rising star on his field. He has a lot of papers on 1st tier vision conference, such as CVPR, ICCV and NIPS. On System area, you can visit IPADS to find a good PI. IPADS have now been publishing papers on OSDI and SOSP, and is becoming the strongest research team on computer systems in China. You may also refer to PhD students at SJTU ACM class, who have collaborations with people from CMU/UW/etc. Then you have somehow started working with a PI from a great US university.
• Year 3: As soon as you get a unique name from UMich, you can start to find a research at UMich. UMich has a lot of research resources, and you should apply broadly to get a position. It is almost impossible to get a position in CV area, as UMich has very few CV faculties. However, other PIs in general ML and System area, some of whom may be outside CSE, might be very open to accept new students. After you officially start your 1st semester at UMich, you will have substantial freedom to take courses and get a very good grade. Use this opportunity to broaden your understanding of CS field, and work on your research to get some publications. In summer, your research advisor may even directly recruit you to be his SURE student, and you can get ~6000$ compensation.
• Year 4: If you have followed my advise, you should have 2-3 publications and probably one 1st author paper. If this happens to you, then you are in a good shape to apply to CS top4 PhD programs. In your application semester, please make your course plan light and focus on your application. Go to your research advisor for strong letters and do not limit the number of programs you choose, unless your research advisor is willing to accept you as his PhD student. Ask people in your field for POI recommendation, and list people you want to work with. Cold-email them if needed and wait for your PhD interview. By the end of February in your final year, you should get a good result.