The AI So-Go-Chi Center will hold its regular research seminar as follows. The seminar will be held in a hybrid format, both on-site and online via Zoom. We look forward to your participation.

Seminar Details

Date & Time: Monday, September 28, 2026, 15:00–16:00 (JST)
Venue: Seminar Room M331, 3rd Floor, Main Building, Research Institute of Electrical Communication (RIEC), Tohoku University / Online (Zoom)
Speaker: Professor Yuko Araki (Graduate School of Information Sciences, Tohoku University)

Title of Talk

Representation and Dynamic Modeling of High-Dimensional Spatiotemporal Data through Functional Data Analysis

Abstract

In many fields, including medicine and the life sciences, cognitive science, engineering, and the social sciences, researchers observe complex data such as curves, images, multichannel signals, and probability distributions, often with temporal and spatial dependence and individual differences. This talk gives an overview of functional data analysis, which treats the continuous structure underlying discrete observations as functions, from the perspectives of representation, dimension reduction, and model selection. It also describes how statistical methods have been developed from applied problems through collaborative research on topics such as gait, gene expression, CT and MRI images, EEG, and the relationship between health indicators and survival prognosis. The talk then introduces recent work on dynamic statistical modeling, including dynamic principal component analysis, distribution-valued time series in Wasserstein space, and the representation of temporal evolution with Koopman operators. Finally, it presents the concept of “AI of AI,” which aims to help researchers choose appropriate models and evaluation methods by connecting statistical science, machine learning, and domain-specific knowledge according to the nature of the data and the research objectives. The speaker hopes to share statistical challenges common to data from different fields and to discuss points of contact that could lead to future collaborative research.