Research Interests

Our research interest broadly lies in statistics and data science. We seek to develop decent statistical methods for analyzing high-dimensional and complex data. We also aim to develop fundamental statistical theories for understanding these methods. The following are some particular areas of our interest.

  • Frequentist theory for Bayes procedures

  • Deep learning from a statistical viewpoint

  • Scalable Bayesian computation

  • Inference in hierarchical latent variable models

We try to keep a balance between theoretical and applied research. Most of our applied work is motivated by collaborations with scientists in various fields of academia and industry. The following are some notable recent collaborations.

  • Collaboration with Samsung C&T and Minseok Song: An important goal of the project is to develop a fashion recommendation system for SSF Shop. We utilized various data such as the click history of SSF Shop visitors, images of fashion items, and a large set of compatible fashion items collected by fashion experts. Some of our research has been reported in news articles.

  • Collaboration with Young-Jin Kim, Young Myoung Ko and Dae-Hyun Choi: This is a collaboration with an industrial (Young Myoung) and electrical (Young-Jin and Dae-Hyun) engineers, aiming to develop an optimal operation strategy for next-generation distribution networks. Our primary role in this project is to quantify the uncertainty in predicting the demand and production of electric power. An ultimate goal is to formulate the problem as a distributionally robust optimization and develop an efficient algorithm to solve it in real-time.

  • Collaboration with Yogiyo and Dong Gu Choi: Food delivery companies in South Korea have rapidly grown in the last decade. There are a lot of intriguing data science problems for food delivery services, and we are working closely with Yogiyo, one of the largest food delivery companies, to find and solve those problems. The present issue is to predict the demand for food delivery which can be utilized for the dynamic pricing of rider commission. Some of our research has been reported in news articles.