In the May 2023 Commencement Celebration, PhD student Yicong Jiang was awarded the 2022-2023 Dempster Prize for his paper co-authored with Professor Tracy Ke, "Semi-Supervised Community Detection via Structural Similarity Metrics." The Dempster Prize is named in honor of Emeritus Professor Arthur P. Dempster and is given annually to a graduate student in recognition of the student’s research. In the following excerpted and edited interview, Yicong shares with us how the Stats Department helped him adjust to living in a new country, discover his passion for research, and hone his ability to collaborate with other researchers. Congrats to Yicong!
1. When did you first become interested in statistics?
Jiang: When I was in elementary school, my parents sent me to an additional school to learn higher level math, which is a common practice among parents in China to ensure that their kids get into a good high school. Students who take part in competitions, e.g. math competitions, and receive awards, are often assigned to a better, more competitive high school than the normal random assignment to a school. After high school, I went to Peking University, where I was a math major, which included both pure math and statistics. On the one hand, I really loved the intrinsic relationship between numbers in pure math. However, during my sophomore year, I realized that I was also interested in statistics because it reveals the relationship between data and the latent structure of data. Statistics helps us to estimate certain parameters or to discover structures like representations. I continued to be intrigued by statistics when I arrived at Harvard for my PhD because, for the first time, I could dive into research to gain novel insights into the relationship between data.
2. What have been some highlights of the PhD program? Are there certain mentors, courses, or parts of the program that really shaped your experience and interest in the field?
Jiang: As I mentioned, after undergrad, I came to the US for the first time to pursue my PhD at Harvard; I was excited to study at one of the best universities for statistics. While I was a little nervous at the beginning of my program, I soon realized that the people here – my advisors, professors, and cohort – were very friendly and helpful. Because the people in the department helped me to learn information about the PhD program, Harvard, and the US in general, I adapted to my new environment in about 1-2 months, and now, I’ve been here for three years!
From the beginning of my PhD experience, Stat 210 [Probability I] stands out as a course. As you may know, Professor Blitzstein is a superstar on YouTube! In his 210 class, he did an excellent job of teaching abstract, difficult concepts by using simple examples and stories. In doing so, Prof. Joe brought abstract concepts of probability close to us – we could almost reach out and touch them.
Another influential course for me was Stat 303 [The Art and Practice of Teaching Statistics], which I took with Professor Pragya Sur. By teaching us research communication and presentation skills, Prof. Pragya’s course helped my cohort build a collaborative research community. Before taking the course, we mostly worked independently on projects without seeking much feedback, but through the course, we learned how to effectively discuss and exchange research ideas with each other. Also, the course helped me to learn professional communication skills in English, including how to highlight core ideas in my presentations instead of simply showing slides with many details and formulas.
3. How did you select your research project? What questions were you asking?
Jiang: After I read Professor Tracy Ke’s paper on community detection in an unsupervised setting (meaning that you don’t have information about the data), I was interested in her methods for clustering different people into groups and wanted to explore this concept with her. The paper that I wrote with Prof. Tracy, “Semi-supervised Community Detection via Structural Similarity Metrics” (semi-supervised means that you have a limited amount of information about the data), uses statistical methods to draw conclusions about communities within a network, such as a social media network. For example, let’s say that you are interested in clustering people into political groups, Republicans and Democrats, but you only have limited information about the politicians that they follow on Twitter [now X]. If “Alice” and “Bob” follow most of the same politicians, we might expect that they are part of the same political affiliation. But how do we measure or score the closeness of their relationship to Democratic or Republican parties? Using structural similarity metrics, we can calculate the degree to which Alice and Bob exhibit the same following behavior and are likely to be a part of the same political community. These methods for community detection can be applied in other ways, such as making recommendations of artists to people based on an artist they follow or making recommendations of people to connect with based on their shared passions.
4. What are you looking forward to? What are your short-term goals?
Jiang: Prof. Tracy and I want to continue to work on semi-supervised community detection to develop methods for addressing the scenario in which people don’t just fit one category or another. Within our current framework, we only make the judgment that people are pure Democrats or pure Republicans, for example, but sometimes people might be neutral to varying degrees (e.g., they might identify as 30% Republican leaning and 70% Democratic leaning). Because there are many possible percentage combinations where people can be considered neutral, the problem is quite difficult.
Outside of research, I am looking forward to a road trip to Yellowstone Park that my friends invited me to join (we are still in the planning phase). Our plan is to drive there; I know this sounds a little crazy, but my friends want to see the sights along the way from Boston to Yellowstone. I’m excited to go hiking because it was an activity that I enjoyed doing with my father when I was younger, but I’m also a little nervous because I haven’t had time to go hiking that much since high school. As long as I have sturdy shoes, I should be fine!