Abstract: A graph consists of sets of nodes and edges and is a flexible object to represent intricate relationships between variables. Examples of graphs can be found in 3D mesh, brain connectivity and molecular structure, and even a simple 2D image can be considered as a graph where individual pixels are uniformly connected with neighboring pixels. Investigating patterns in graphs requires sophisticated machine learning (ML) methods as the connections between the nodes are typically heterogeneous, and learning on graphs has become a fundamental topic in Artificial Intelligence. In this seminar, I will cover a variety of graph analyses methods ranging from statistical methods to deep learning, under the theme of multi-resolution in traditional signal processing. The multi-resolution graph ML methods introduced here will mainly address problems in identifying Alzheimer’s Disease (AD) specific variations with structural brain connectivity from diffusion weighted images (IPMI 2021, MICCAI 2022 and AAAI 2024) as well as how the multi-resolution scheme helps generate real-like synthetic graphs on standard graph benchmarks in ML (NeurIPS 2023).
Brief Bio: Won Hwa Kim is an Associate Professor in Graduate School of Artificial Intelligence (GSAI) / Computer Science and Engineering (CSE) / Medical Science and Engineering (MED) at Pohang University of Science and Technology (POSTECH). Prior to joining POSTECH, he was an Assistant Professor (tenure-track) in Computer Science and Engineering at the University of Texas at Arlington (2018 – 2023, last 2 years on leave-of-absence), and he was a Researcher in Data Science Team at NEC Labs., America (2017-2018). He obtained a PhD in Computer Sciences from University of Wisconsin-Madison in 2017, an MS in Robotics from KAIST (2010) and a BS in Electrical Engineering from Sungkyunkwan University (2008). He developed Hybrid Vehicles at Hyundai Motors Company in 2010-2011 before he dived into Artificial Intelligence.
