O Speaker: Prof. Jae Kyoung Kim

O Affiliation: Mathemtical Sciences & IBS
                  Biomedical Mathematics Group

                  KAIST

O Date: March 05, 2025

O Start Time: 4:00 PM

 

O Abstract: 

In this talk, I will present methods for extracting meaningful insights from both static and time-series data. For static data, Principal Component Analysis (PCA) is a common tool for identifying signals in noisy datasets. However, selecting the optimal number of signals often involves subjective judgment. To address this, I will introduce a novel approach based on random matrix theory, enabling an objective determination of the optimal signal count, demonstrated through its application to single-cell RNA sequencing (scRNA-seq) data. For time-series data, a range of statistical and machine learning-based methods have been developed. I will show how combining these approaches with mathematical modeling can significantly enhance their predictive power. This will be illustrated through a new algorithm that accurately predicts mood episodes (depression and mania) using only sleep-wake data collected from smartwatches. By integrating advanced mathematical techniques with real-world applications, this talk aims to highlight innovative strategies for tackling complex data challenges.

 

 

* The seminar will be held offline and online [https://kaist.zoom.us/j/86770940131] simultaneously.

 

* Closed Caption interpretation service (English) available online (service provided by Zoom)