Alumni Q&A on Maximizing Your Time in ACME to Prepare for Grad School (Part 2)
A guest article written by Connor Robertson
In the last article, I shared my view on how to maximize your time in ACME to prepare for grad school. Since all ACME alums have different experiences in graduate school, I wanted to include Q&As that give perspectives aside from my own. The following are different alums’ views on their experience in graduate school:
Jacob Heiner – Class of 2018
Q: What academic experience have you had since ACME?
A: Since ACME, I have completed a one-year master’s degree at the University of Washington.
Q: What was the topic of your research or projects?
A: I elected to do a no-thesis option, as I determined that I was not interested in continuing in academia. I did a handful of small projects using machine learning to discover governing equations of physical systems, which was very interesting.
Q: What ACME material did you use in your projects?
A: In those projects, my skills using Python were invaluable. I was also grateful for the linear algebra education provided by the ACME program. I could have skipped the first PhD level linear algebra class and been fine.
Q: What do you wish you had done during your time in ACME to prepare more?
A: One thing I wish I had done prior to graduate school was figuring out what I wanted out of a graduate program. There are a lot of paths, and had I known what I wanted to do, I could have chosen classes that were much more relevant to what I wanted to do and that educational experience would have been much more valuable to me.
Q: What has surprised you most about graduate school?
A: I was surprised by the practical focus of my degree program. Much of the learning was for the purpose of solving problems, where I had expected it to be more rigorous and focused on proving theorems like ACME.
McKell Woodland (Stauffer) – Class of 2018
Q: What academic experience have you had since ACME?
A: A year of graduate work in CS at BYU, a year of graduate work in CS at Rice while researching at MD Anderson Cancer Center. I also did summer research at Lawrence Livermore National Labs & Texas Children’s Hospital & Microsoft Security.
Q: What was the topic of your research or projects?
A: BYU: Integrating representation learning into Deep reinforcement learning.
Lawrence Livermore: Integrated a R2U-Net into a ConvLSTM for an event detection system in video.
Texas Children’s Hospital: Built a junctional ectopic tachycardia detector for postoperative cardiac surgery infants.
Rice/M.D Anderson: Working on automating the process of segmenting CT scans so that an updated deformable model of the infected organ can be given to the doctor right before radiation therapy.
Microsoft Security Research: Building a process to detect overlapping suspicious security threat instances so that Microsoft Threat Experts have a smaller and more informative alert base to look at.
Q: What ACME material did you use in your projects?
A: Honestly, I don’t use what I learned in ACME much as my research has focused on computer vision aspects of neural networks. That being said, the mathematics behind optimization helps me to understand neural network research. ACME helped prepare me to pick up new sciences & technologies fast.
Q: What do you feel sets you apart from your current classmates as an ACME alum?
A: I honestly have felt no advantage at Rice, as the majority of the students in my program have a strong mathematics and computer science background from prestigious universities around the world. But I think ACME gave me the ability to get into and excel in highly rigorous situations. I feel totally comfortable and on par with my classmates.
Outside of academics, the major difference between my classmates & I is a work life balance. For my own sanity, ACME taught me to eat right, sleep, work out & prioritize time with family and my spiritual life. It was so hard that it taught me not to be perfectionistic and how to study efficiently and take care of myself so I could be done. My poor classmates have never learned this. Many of my officemates hardly ever leave the office & are always talking about how stressed/depressed/tired/overwhelmed they are.
Q: What has surprised you most about graduate school?
A: I think the biggest surprise was the change in dynamics. ACME is very time consuming and you go very fast through the material. On the other hand, graduate work goes very deep, the classes don’t have much work associated with them, and you can make research as time consuming as you want it to be. Instead of having a structure, you have to learn the self discipline to work hard even though there are no immediate consequences for not. I honestly struggled for a while being stripped of structure and learning to push myself to be in charge of my own learning. It has also been an interesting transition to learn how to integrate everything I know to solve a problem, instead of just focusing on current concepts.
Kevin Miller – Class of 2016
Q: What academic experience have you had since ACME?
A: I went straight into my PhD in Mathematics at UCLA, though I’ve been awarded a Master’s Degree along the way. I have been supported by an NSF NRT Graduate Fellowship (2018-2019) and I am currently supported by the DOD National Defense Science and Engineering Graduate (NDSEG) Fellowship through the rest of my PhD.
Q: What was the topic of your research or projects?
A: I am looking into what is called Active Learning, basically making decisions amidst uncertainty for application in graph-based semi-supervised learning. I have previously worked with community detection and link prediction in network science, as well as worked on proving posterior consistency results for graph-based semi-supervised learning.
Q: What ACME material did you use in your projects?
A: Throughout all of my research, the understanding of analysis from Volume 1 has been very influential, as well as the coding and optimization background that I gained from Volume 2. The probability foundation I got in Volume 3 has also been a great preparation for my further research in machine learning. To be more specific, my work with posterior consistency and network science has relied heavily on eigenvalue and eigenvector analysis, and so the foundational analysis principles in Volume 1 were used a lot. Then, since all of my research has fallen under the broader umbrella of machine learning applications, the coding skills and nonlinear optimization background I gained in Volume 2 of ACME has been a great foundation. Most of my current research problem of Active Learning really is trying to define a well-principled optimization problem that I can solve efficiently in order to make decisions under uncertainty.
Q: What do you wish you had done during your time in ACME to prepare more?
A: I wish I had done more to synthesize the topics I was learning in the homeworks and chapters. At different times of my time in ACME I think I was doing better at really understanding the material and not just getting the homework done as fast as I could in order to make a deadline. I think sometimes with the big workload, it’s easy to get overwhelmed and lose sight of the big picture of what you’re really learning and why the details of the proofs are important. The opportunity and materials are all there in ACME, it’s just making sure to be disciplined enough to take full advantage of them.
Q: What do you feel sets you apart from your current classmates as an ACME alum?
A: The breadth of experience and topics we receive in ACME far surpasses what others in my program have done. Each graduate class I’ve taken at least started with topics and principles we talked about in ACME. For most mathematics undergraduates, they have to choose specific classes to be able to get any exposure to the wide variety of topics in applied mathematics out there – ACME provided a good exposure to them all. Also, I think the cohort-style plus integration of coding/lab experience with the lectures provided me with a better understanding than others about how this all actually is applied in real-life. Others in my current program have had to learn how to code and understand how our work actually is applied, while ACME provided that experience naturally because of the way it is designed.
Q: What has surprised you most about graduate school?
A: Graduate school has been much more individual in its experience. You create your experience, it is not just handed to you like it is in ACME. This has been good and bad at times for me. I have enjoyed blazing my own trail if you will, choosing what research experiences and classes I want to follow. However, it has been a little bit lonely at times, because my interests have been different than others in my program. It is great to have the opportunity to grow individually and learn deeply about research topics with some of the brightest people I’ve ever met, but often “solo” experience can be difficult to handle depending on the type of program you go into (e.g. “big” vs “small” program).
Shane McQuarrie – Class of 2016
Q: What academic experience have you had since ACME?
A: I got a Masters degree in Math from BYU after finishing ACME, then went to the Oden Institute at the University of Texas at Austin for a Ph.D. in Computational Science, Engineering, and Mathematics (that’s where I am now). I also did one summer internship at Sandia National Labs during my Masters degree. As a senior and during my Masters degree, I was an ACME TA and worked a ton on the lab curriculum.
Q: What was the topic of your research or projects?
A: At BYU I worked on a data assimilation problem for a fluid flow problem (Boussinesq equations). At UT Austin, I’m working on data-driven, nonintrusive reduced-order modeling for large dynamical systems (combustion, additive manufacturing, potentially plasma torches)
(translation: rocket science, 3d printing, and lightsabers!)
Q: What ACME material did you use in your projects?
A: For both the Masters degree and my Ph.D., the programming aspect has been hugely important, especially
- NumPy / SciPy for computational work, duh
- Data visualization principles and plotting tools
- Learning the unix shell, which was essential to know on the way to supercomputing
- Knowing Jupyter notebook, it’s a lifesaver for research!
- Good coding principles (writing readable code, profiling, unit tests, etc.)
- Knowing how to do essentially anything else in Python, not just numerical work. It’s super helpful for file organization, parsing data files, etc.
- On the mathematical side, I’ve really needed
- Linear algebra (thanks, Volume 1!)
- Optimization techniques (thanks, Volume 2!)
- Machine learning algorithms and techniques (thanks, Volume 3!)
- ODEs / PDEs, theoretically and numerically (thanks, Volume 4!)
I’m serious, I’ve absolutely needed all of these things in my work. They all pop up everywhere, all the time.
Q: What do you wish you had done during your time in ACME to prepare more?
A: Maybe learned a little about ACME-related graduate school opportunities or gotten exposed to other research problems that aren’t studied at BYU. I had never even heard of model reduction until I got to UT Austin, and now that’s the main focus of my research. It could have made it easier to find graduate programs that I might be interested in later on.
Q: What do you feel sets you apart from your current classmates as an ACME alum?
A: The students in my current Ph.D. cohort work on a wide variety of problems––electrodynamics, computational oncology, finite element analysis, Bayesian inverse problems, advanced materials science, high-performance linear algebra––and while I’m not an expert on the problem domains, I can talk to literally any of my classmates about their research and understand the main ideas of what they’re doing. I also definitely had a huge advantage in programming and mathematical analysis going into the Ph.D. curriculum.
Q: What has surprised you most about graduate school?
A: I was honestly shocked by how well prepared I was for my Ph.D. program. Not because I’m a genius or because my classmates are dumb (they are the smartest people I know!), but because of the sheer volume of topics that we covered in ACME. I was familiar with at least 85% of the topics that we covered in my graduate curriculum––and we were not simply repeating ACME topics. I have some classmates that are math geniuses and others that are master coders, but I think I had the strongest balance between the two areas.
Mitchell Sailsbery – Class of 2018
Q: What academic experience have you had since ACME?
A: PhD program at University at Buffalo (Presidential Fellowship, Engineering Design Research Assistant, EDGE Fellowship/Internship). Left Early with Master’s.
Q: What was the topic of your research or projects?
A: Mean Field Games and Reinforcement Learning.
Other School Related Research Projects: Multi-Agent Reinforcement Learning Through Mean Field Games, Solomonic Bargaining in Legal Property Rules and Social Welfare, Generation of Foams for Manufacturing Using Mean Curvature Flow, Topological Data Analysis Through Estimation of Varieties, Stabilization of GANs through Mean Field Games, Resource Allocation in Cloud Computing Using Network Topology.
Q: What ACME material did you use in your projects?
A: In order of magnitude, from largest to smallest: Linear Algebra, General Analysis, Probability Theory, Coding Workflow, Algorithm Analysis, Numerical Analysis, Control Theory, General Machine Learning and Statistics, Partial Differential Equations.
Q: What do you wish you had done during your time in ACME to prepare more?
A: In general, PhD programs do not award generalists. If you know who you want to work with before applying, you can get a good head start on earning a PhD efficiently.
Q: What do you feel sets you apart from your current classmates as an ACME alum?
A: General applied math background, Coding skills and workflow, Comfort self-learning and making good use of learning in groups.
Q: What has surprised you most about graduate school?
A: Having very specific academic goals is well-rewarded. Also, being a self starter (seeking out study groups, mapping out and planning degree progress, reading lots of papers early on in the program, etc.) is expected.
Matt Schaelling – Class of 2018
Q: What academic experience have you had since ACME?
A: Two years of a predoctoral fellowship through Stanford Institute for Economic Policy Research.
Q: What was the topic of your research or projects?
A: I worked in Susan Athey’s lab on projects related to machine learning and causal inference. Here, I assisted in leveraging new machine learning methods in empirical projects relating to student financial aid, voter participation, charitable giving, and retirement savings. These methods, developed within the lab, adapt the random forest model to estimate heterogeneous (or “personalized”) treatment effects for policies/interventions. This differs from traditional machine learning methods that focus on predicting an observed outcome (y), whereas we are estimating an unobserved quantity (the treatment effect).
Q: What ACME material did you use in your projects?
A: General computational theory has always been useful (e.g. thinking about complexity, writing performative code for numerical computation). One of the first projects I worked on used information theory — the estimation algorithm focused on minimizing the Kullback-Leibler divergence.
Q: What do you wish you had done during your time in ACME to prepare more?
A: I wish I would have spent more time on my research assistant work. It was easy to feel the allure of all the sexy stuff that ACME focuses on preparing students for, like machine learning, in such a way that caused me to overlook really interesting research happening in other fields/departments. Maybe there was an economics professor who did research that I would have really enjoyed working on, but because it didn’t sound like it used enough complicated math or sophisticated enough models. All that is to say, focus on optimizing your own objective function, not someone else’s.
Q: What do you feel sets you apart from your current classmates as an ACME alum?
A: I’m not afraid of difficult math.
Q: What has surprised you most about graduate school?
A: While doing ACME definitely prepared me a lot, it quickly became apparent to me that others were super well prepared for graduate school too. Don’t expect things to be dramatically easier for you just because you did ACME.
Connor Robertson – Class of 2018
Q: What academic experience have you had since ACME?
A: I went straight into a PhD in Applied Math at the New Jersey Institute of Technology.
Q: What was the topic of your research or projects?
A: My research focuses on deriving the governing equations of physical systems directly from sensor or image data.
Q: What ACME material did you use in your projects?
A: I have used pieces of all 4 volumes: linear algebra, interpolation, numerical analysis, modeling with PDEs, and probability.
Q: What do you wish you had done during your time in ACME to prepare more?
A: I wish I had explored more topics that I might be interested in researching. I never would have imagined what I am currently working on and it took trying a few different projects before I found it. The other projects were informative and interesting, but ultimately, I think I could have honed in more on what I enjoy if I had just spent a bit more time looking around.
Also, I should have payed closer attention to all the modeling and differential equations. My current program is full of professors who specialize in those areas and it is the only place I’ve been behind my classmates in my classes.
Q: What do you feel sets you apart from your current classmates as an ACME alum?
A: Exposure to topics and computer science knowledge and skills. I think I had the broadest exposure to topics of my classmates and I have yet to take a class that doesn’t start with or include something we went over at least briefly in ACME.
On the programming side, I think I have had a big leg up in progressing my research because of all the practice we had implementing algorithms in the ACME labs. Also, the extra computer science courses I took and working on the ACME Dev team has helped me quickly get going with supercomputing.
Q: What has surprised you most about graduate school?
A: It takes a lot more effort in graduate school to work with other people. ACME makes it easy to form a group and work with your classmates because you are all spending so much time on the exact same material. But in graduate school, you have to get outside your shell and talk with your classmates until you find people you can bounce ideas off of. However, the effort has definitely been worth it because I definitely learn faster in a group.
It’s also surprised me how important it is to choose your advisor. Not only based on what they work on and if it’s interesting to you, but also their style of working. For a PhD, you are usually looking at a minimum of 4 years working closely with that person and if you can’t communicate well or align your ideas, it is going to be a really rough time. Older graduate students can give good insights what professors work on and their working style.