-
MEMBERS ※As of April 2024 (Click here for previous members)
Department of Physics and Mathematics
Assistant Professor Yuichiro Kobayashi (Representative)
-
OVERVIEW
Recent developments of machine learning techniques allow highly precise predictions based on a large amount of data, commonly characterized by a large number of variables for a single data (such as an image). The developments are particularly remarkable and visible for data with prescribed formats, such as images, audio, and texts. However, rather complex data are usual, for example, for economic phenomena. For a business firm, time-series, categorical, geographic, natural-language, and network-type data might all be available. By focusing on the general competence of networks (i.e., graphs) to represent diverse types of data, our project is aimed at building a unified framework whereby data with a complex structure could be analyzed systematically and with interpretability.