This Q&A is divided into four main categories: Research, Advising, Funding, and Lab Culture. They are tailored to PhD students but can be relevant to Master’s students as well.
What type of research do you like to do? Theoretical vs applied? Software vs hardware?
I like to stay somewhere in the middle on the “theory vs application” spectrum. I have a hard time getting excited about something unless I can see some (mathematical) abstraction that explains why it works that way, but I also have a hard time staying excited for a long time unless I can see it working in a realistic application. I expect that each student will have their own preference for where to be on these spectra. The main thing from my perspective as an advisor is that regardless of your specific interests, you should have a strong background in mathematics and be comfortable connecting (your own and others’) abstract research ideas to (your own and others’) more concrete applications.
What types of classes should I be thinking about taking? Any other topics you’d recommend I read up on in my first year?
You are in the driver’s seat here, so at the end of the day it’s up to you! That said, I will strongly recommend that you become proficient in the following material, either by taking classes or by self-study: optimization (not just convex!), probability and stochastic processes, linear and nonlinear system theory. If you are interested in multi-agent systems, you should plan to take my graduate course on game theory in your second year (it presumes familiarity with optimization).
Note that I have very explicitly not recommended that you jump right into a course on machine learning or reinforcement learning. My experience is that taking these courses before nailing down your fundamentals in optimization and control theory can be inefficient. If you are interested in these topics, I certainly encourage you to take courses on them, but you will get more out of those courses if you are already familiar with the fundamentals. Walk before you run.
Besides classes, there are a lot of other really great resources out there you should be familiar with. Some particular favorites of mine: the matrix cookbook, MIT OpenCourseWare, 5 minutes with Cyrill, and Maths for Intelligent Systems.
What kind of progress do you expect from me?
Once you have (essentially) finished courses in your second year, you should be able to spend full cycles on research. In your first year, I expect that you will focus on taking classes, building your ability to read and present academic papers, and start to get your feet wet with research by working on an existing (or new!) research project in the lab. The first year will look different for different people, but the bottom line is that I expect that by the end of your first year, you should have mastered most of the essential skills you need to engage with research and passed your written qualifying exam. A good goal for all first year students is to submit a first-authored conference paper in the fall of second year (e.g., ICRA, ACC).
In subsequent years, I expect that you focus almost all work-related effort on research and at any given time maintain a mix of projects you are “leading” and those you are “helping out with.” Every student’s journey will be different, though, so if you are not sure how things are going or how to plan your schedule, let’s talk.
How many papers you expect from your students? What is the typical lifecycle of a project/paper?
1-2 a year is a decent number. You should be able to hit that with absolutely no problem by your third year. More is better, if they are high quality. It is worse if they are getting in the way of you being able to think deeply about more significant research problems. By your third year, you should be able to go from “rough idea” to “submitted conference paper” in roughly 6-9 months. Obviously, this is something that you get more comfortable with over time and as you get more experienced. So, for example, I expect most people’s first paper to be based on work they start by the end of the first semester, and not get submitted until the end of the following summer. By your fifth year, you should probably be able to work on 3-4 projects every year, or more.
What kind of projects do you see me working on?
I want them to be mostly coming from you, and adapted based on our discussions. But you will be the driver. I hope that your projects will teach me new things!
When working on new ideas, how do you know when it looks like a dead end, and how do you determine that?
I think the secret is to just have multiple projects going on at any one time. Your first project is usually going to be something that I help suggest… and therefore it’ll almost certainly be “low risk” from a technical perspective. In future, you should have multiple of “your own” things going on at once so that you can bounce between them as you feel one of them hitting a snag, then come back when you want to.
How do you recommend that students stay motivated and disciplined?
Keep a normal 8-5 schedule. Don’t work too hard on the weekend (maybe just 4 hrs on a typical weekend once you have finished taking all your classes and are in “research-only” mode, more before then). Make sure you have multiple projects, with multiple different collaborators. That helps keep things exciting.
How do you recommend balancing research with other commitments such as classes, teaching, and quals?
Research should be your highest priority at the end of the day. The whole point of a PhD is to learn to be an independent technical thinker, and research is the way you build that skill. At times some commitments will take temporary priority, e.g., you need to make sure you study hard for quals a couple weeks before the exam in order to make sure you pass. However, it is important to keep your research progress in mind. For example, do not fall into the trap of automatically prioritizing homework assignments because they have a due date and research does not.
Grades are a bit less meaningful in graduate classes than in undergraduate classes; typically, grad classes are graded very generously and are not intended to eat too much time that could be used for research. As a rough guideline, anything lower than an A- should start to sound some alarm bells.
That said, you should absolutely be attending all of your classes and spending the time you need with the material in order to understand it. After all, the point of classes is to learn the material, not to spend the minimum amount of time in order to get a decent grade. In your first year, this will mean you have a bit less time for research. However, you should make sure that you regularly still carve out at least some significant blocks of time to do research.
A final note: if you have difficulty balancing research with other commitments such as classes, do not be surprised if I recommend that you TA at least part time while you are finishing your classes. This will allow you to gain valuable experience presenting technical material during the semesters in which classes are impeding your ability to prioritize research. In this situation, we will discuss expectations regarding research productivity on a case-by-case basis.
I just got reviews back on my paper. Now what should I do?
First things first: it is very tempting to look at a negative review and think “Oh man, those reviewers were totally out to lunch and they missed the whole point of the paper because they were so incompetent. I can totally disregard what they have to say.” NO! Do not do this! Reviewers are doing a public service by reading our papers, and should always be thought of—not just treated—with respect. Even when you get a review that you don’t agree with or that demonstrates an egregious misunderstanding of the paper, you should think to yourself: “Ok. The reviewer is a very smart person who might have been reading this on a plane or in a hurry. They misunderstood something important about my ideas. How can I present the ideas more clearly to minimize the chance of a reader doing the same thing.” Of course, sometimes reviewers point out serious technical flaws; again, take these seriously and do not ignore or whitewash these issues. If it means removing an improper result or even retracting the submission, do it—better to find out before publication than afterward.
Another perspective: when I review papers, part of me is trying to learn something new about the topic at hand, but part of me is also trying to save the authors from potential future embarassment, e.g., from incorrect results, poor literature review, theoretical issues, bad writing, …. Interpret reviewer comments accordingly.
Now, if you need to prepare a rebuttal, a few pieces of advice:
How do you prefer students to communicate with you?
Effective communication is absolutely essential for us to work together well. Typically, we will have a 1-1 meeting for 30 minutes every week, although this may vary on a case-by-case basis. I expect that during this meeting we will check in on technical, administrative, and general life updates. When I was a PhD student, this was usually the high point of my week, and I hope it can be for you as well.
A few other things to note: (1) I try to be very responsive on email and on Slack, but I find that these media are not particularly efficient for conveying technical ideas and are best used for quick messages like “I can’t make our meeting tomorrow” or “Can you take a look at my latest paper draft?” (2) While communication is good, too much communication can become inefficient and tiring. Please try to keep our communication channels clear so that each message remains important. (3) Relatedly, “ghosting” is never appropriate. I will not ghost you, and I expect you not to ghost me.
What do you do when students are struggling? How do you know when to step in and help more vs. letting the student figure it out?
I err on the side of letting students figure things out themselves, with only minimal technical pointers from me. A PhD is not about being taught by an advisor, at least not in the sense of learning lots of technical details from him/her. It’s about being advised to investigate interesting questions, and learning to find the resources and technical capabilities you need to figure them out. An advisor can help point, but is not the person who teaches those capabilities, at least not most of the time.
Put another way: if a student will always be struggling without their advisor’s support, then the advisor has failed to train a self-reliant student who will be capable of success in the future. You are in charge of your own education, and I expect you to take the lead!
How do you approach the writing process with students?
Communication is probably the most important part of science and engineering. I expect students to develop excellent communication skills, both in writing and in speaking. I will happily look at (a bounded number of reasonably polished) drafts and give feedback, but you need to be proactive in writing early, soliciting feedback, revising, and repeating the loop as much as needed until you have a manuscript you are proud of. Please consult the lab writing guide for more details. For advice on giving talks, I strongly recommend watching Talks that Don’t Suck from Cyrill Stachniss.
For collaborative projects, what is the expected level of contribution from everyone for the collaboration to work?
It varies from project to project, and should be organic in my opinion. People should contribute as much as they want to, and if no one wants to contribute enough to make the project work, then the project was always going to be doomed. If someone is not contributing meaningfully, then they should politely tell their collaborators that they are either uninterested or do not have bandwidth to be a good collaborator. On the other hand, there can be tremendous value to having a collaborator who only contributes periodically but whose input helps direct the project in useful directions. Contributions come in many forms.
The lab seems to be growing larger pretty quickly. How do you support everyone?
With difficulty and much proposal writing. Also, I expect every student to TA at least two semesters during the PhD. This will vary depending upon funding needs, but TAing is a really important educational experience and provides a great way to practice technical communication. I found it essential as a graduate student.
What do you see me doing during summer? Internships?
Whatever you want, for the most part. There is a chance I might prefer you to intern certain summers if I am short $$. The sooner we talk about things the better.
Would you help students enter the job market?
Definitely!
Do you anticipate funding to change during my time as a student?
Yes, the source will definitely change. I will do my best to make sure it does not affect your research at all. However, I may ask you to attend meetings and present your work to sponsors; I did this many times as a PhD student and it can be a great way to get some face time with important people, meet potential collaborators, etc.
Should I apply for fellowships?
If you are eligible and have good grades, yes. A (partial) list of fellowships is: NSF GRFP, NASA NSTGRO, DOD NDSEG, DOD SMART, Hertz, Siebel, Microsoft, Google, NVIDIA, Apple. For a more comprehensive list, you can check out the one maintained by CMU.
The lab is still relatively new; what are your plans for the lab for the future?
I want to build a vibrant lab where ~6-10 PhD students are intellectually curious and self-motivated to study questions they find exciting at the intersection of optimization, control, games, etc. I want to learn new things from my students, and I want everyone to enjoy working together.
How much are we expected to work on papers alone vs. with other colleagues/students? How would we compartmentalize tasks?
You should have something — a “story” — that is your own, primarily. This will be your thesis. On the other hand, that “story” can and should intersect with others’ and you should collaborate to your heart’s content. For example, my “stor” was fast algorithms for N-player smooth dynamic games, with applications to human-robot interaction. I was the primary driver behind the algorithmic part for games, but literally every single paper had at least one other student coauthor, and in half/most I was not the first author. This was great. I would have quit the PhD if it had all been solo.
What about travel and vacation?
Vacation is important, and I trust you to take vacations when you need them. That said, remember that doing a PhD is about research, and research does not necessarily align with the university’s academic calendar. Be aware of upcoming project deadlines and plan your vacations accordingly. Also, I find that “working on vacation” is rarely as productive as anyone thinks it is going to be… so please think carefully when planning trips. Last, please remember to let me know when you will be out of town and cancel any 1-1 meetings that are scheduled in your absence.