Presentations | Yong-Min Shin

Featured: Seminar Series

Seminar Series on Graph Learning GIST (Gwangju Institute of Science and Technology) · 2025

A series of seminars given as a guest speaker in Prof. Hong Kook Kim's group from March ~ May 2025. The seminar series covers the basics of graph learning, graph neural networks, several fundamental concepts, and applications.

Session 1: Introduction to graph mining and graph neural networks Session 2: On the representational power of graph neural networks Session 3: A graph signal processing viewpoint of graph neural networks Session 4: From label propagation to graph neural networks Session 5: On the problem of oversmoothing and oversquashing Session 6: Towards efficient graph learning Session 7: Explainable graph neural networks

External Presentations

2025

Faithful and Accurate Self-Attention Attribution for MPNNs via the Computation Tree Viewpoint AAAI 2025

Poster for my paper accepted at AAAI 2025. The arxiv version can be found here.

Faithful and Accurate Self-Attention Attribution for MPNNs

A Practical Introduction to (Explainable) Graph Learning Ewha Womans University

A seminar given as a guest lecturer in April 2025.

A practical introduction to (explainable) graph learning

2024

PAGE: Prototype-Based Model-Level Explanations for Graph Neural Networks Best Academic Paper Award, Yonsei University

Presentation slides submitted as official supplementary material after receiving the best academic paper award at the Graduate School of Yonsei University.
Original paper: Shin et al., "PAGE: Prototype-Based Model-Level Explanations for Graph Neural Networks", TPAMI (2024)

PAGE: Prototype-Based Model-Level Explanations for GNNs PAGE (AAAI 2022 short version)

On the Feasibility of Fidelity- for Graph Pruning IJCAI 2024 Workshop on XAI

Poster and presentation slides presented at the Workshop on Explainable Artificial Intelligence (XAI) at IJCAI 2024. The arxiv version can be found here.

On the Feasibility of Fidelity- for Graph Pruning On the Feasibility of Fidelity- for Graph Pruning Accompanying blog post

2022

Edgeless-GNN: Unsupervised Inductive Edgeless Network Embedding 28th Samsung HumanTech Award (Bronze Prize)

Presentation given during the competition for the 28th Samsung HumanTech Award.
Original paper: Shin et al., "Edgeless-GNN: Unsupervised Representation Learning for Edgeless Nodes", TETC (2022)

Edgeless-GNN: Unsupervised Inductive Edgeless Network Embedding

Internal Presentations

In my lab, we have a weekly reading group where we present papers of our own interest. Here are some of the selected presentations that I have made in the reading group.

Causal Learning

A two-part presentation introducing the basics of causal learning. I got a LOT of help from "Introduction to Causal Inference" by Brady Neal. Definitely check out his course if you are interested — I personally had a lot of fun learning from it.

Part 1 Part 2

Self-Supervised Learning

A presentation made when I was interested in understanding the SimCLR framework. Understanding this paper helped me better understand other papers in graph self-supervised learning.

Introduction to SimCLR

Knowledge Distillation

A presentation made when I was interested in understanding knowledge distillation, which I eventually applied to my research on GNN-to-MLP knowledge distillation.

Towards understanding knowledge distillation

Physics-Informed Machine Learning

A presentation made when I was interested in understanding the physics-informed machine learning landscape.

Vector Neurons: A General Framework for SO(3)-Equivariant Networks

Unsupervised Disentanglement

A review of the paper: "Challenging common assumptions in the unsupervised learning of disentangled representations" by Locatello et al. This paper is a great read if you are interested in disentangled representation learning, and personally, I found it very helpful while reading some papers in explainable AI.

Challenging common assumptions in the unsupervised learning of disentangled representations