Interview with Dr. Luke Bornn

Department of Statistics & Actuarial Science

(Computational Statistics and Machine Learning)

Joined SFU in July 2015

The applications of your research range from the environment to sports. Which application affects you personally or motivates you the most?
At the moment, most of my time is focused on sports, primarily soccer and basketball. I’m also interested in problems related to the Earth’s climate. There are many statistical issues that are glossed over by climate scientists – a lot of uncertainty that is ignored – and I think there is a great opportunity for statisticians to add value in that domain.

What type of data occupies your thoughts the most?
The data I work with is primarily tracking data, for example, the locations of basketball players on the court over time, or seals in the ocean, or people trying to detect explosive devices, where you have movement and outcomes. What interests me is trying to figure out which individuals are making the best decisions for the desired outcomes, whether it be which basketball player makes the best passes, which seals take the best route for finding food, or which search and rescue members make the best use of their time in covering ground to reach a survivor.

Are there a lot of people working on tracking data?
Actually, very few people are working on it, and one reason is that although the raw data itself is really big and unstructured, you need a unique combination of skills beyond merely handling large data sets. There are all kinds of statistical issues to consider such as confounding and separating individual effects from group effects; for example, understanding an individual's contribution to a certain outcome when there is already an underlying strategy in place that applies to the whole team. And then you need the modeling skills to model all of these things. So the work requires a skill set that few people have. Further, these data sets are relatively new. Only recently have we been able to track individuals and groups with such a high level of fidelity. Few people are working with this raw data; on the other hand, many people use summaries of this data, e.g., in the NBA it is used for simple statistics such as tracking how far a player ran or their average velocity.

What is the downside of your work?
We often work with massive amounts of unstructured data, so we spend a great deal of time preparing data for analysis; tedious work like processing data, cleaning data, putting it in a database format. Typically, someone who goes into statistical modeling and machine learning doesn't think about that side of the job beforehand; not only will they be doing fancy modeling and prediction, but alongside that there are these early-stage tasks that involve getting a massive data set into a usable form.

You have conducted research in both the U.S. and Canada.  Are Canadian funding levels especially lower in statistics?
While statistics spawned out of mathematics, in recent years the work of many statisticians aligns more closely with computer science, requiring massive computational resources and labor-intensive data processing tasks. However, due to its origins, statistics continues to be funded at similar levels as mathematics, while computer science funding is significantly larger. While this problem is universal across the US and Canada, in the US there has been a much larger funding push in the areas of data science and big data, from which statisticians have finally been able to obtain adequate funding to conduct their research.

Who or what affected your decision to pursue a faculty position?
I loved research right from the beginning. All of my supervisors were passionate about their research and that passion was contagious. I knew right from the beginning that research was what I wanted to do and the best path to do that was pursuing a faculty position.

How do your former supervisors influence your current training philosophy?
My supervisors allowed me to be self-guided, which was a good match for my character. As a supervisor, I try to encourage students’ independence. When a student starts off you have to give them a lot more direction to establish where their program is headed, but by the end of their training my goal is essentially to watch at a distance. My supervisors did a wonderful job of that.

I believe that you can do a couple of important things for your students. First, you can instill in them a real passion for the subject so that they love the work they do. Second, you can close off dead-ends for them, keeping them on a fruitful path. If you don't have that first piece – the passion – in place, you will need to micromanage everything because they won’t take the work to the next level. But once they have passion for their research, you just need to keep them pointed in the right direction.

What educational background and personal strengths do you look for from prospective group members?
I look for someone who is independent and self-motivated, hard-working. And someone who when given an idea, doesn’t just to do the one thing but instead pushes it several steps further, even if that might not be the best approach; thinking beyond the basic instructions is critical. People focus too much on the technical details, such as whether a student has the right math background, which is less important to me than their work ethic and level of motivation. I've seen technically proficient students come into a research group where they are expected to work on their own, but the skills required for research are not well aligned with the skills required for taking tests at the undergraduate level.

What contemporary scientific issue concerns you the most and requires immediate attention?
Something that concerns me is the entrance of unqualified people into the field of data science. Statistics, in my opinion, is the data science, yet the term has been adopted by people with a wide variety of training. There are  strengths that come with statistical training – an understanding of foundational issues like confounding, design, collinearity, and causality, for example. There are many subtle issues in data analyses that can lead to major issues and major biases if they aren’t considered carefully beforehand; these things are taught in a formal statistics education. It's very easy to take statistical data and come to conclusions with it, but the understanding that comes with a proper statistical education can prevent you from making wrong conclusions.

______________________

Dr. Bornn’s presence enhances SFU’s expertise in the statistical analysis and modeling of big data. His interests complement those of his SFU colleagues in Math and Statistics, Biology, Kinesiology, Health, Computing, and Earth Sciences. His research will harness the immense power of using raw tracking data to make accurate predictions.

Read more: Dr. Bornn’s personal website and the New Science Faculty page

Interview by Jacqueline Watson with Theresa Kitos