12:30 - 13:00 Registration
13:00 - 14:00 Jeong Rye Park
14:00- 15:00 Sang June Lee
15:00 - 16:00 Eunmi Kim
16:00 - 17:00 Jaeyoon Ahn
17:00 - 18:00 Yungbum Jung
18:20 - Banquet
Jaeyoon Ahn
Title : Interpretability of Attention Weights
Yungbum Jung
Title : What is AI semi-conductor?
Eunmi Kim
Title : Link prediction using graph neural network: reconstructing contact tracing network for epidemic simulations
Abstract: Link prediction is an important application of graph neural networks. By predicting missing or future links between pairs of nodes, link prediction is widely used in social networks, biological networks, recommender systems, etc. Graph neural networks (GNN) have been shown to outperform traditional methods for link prediction as a powerful tool for jointly learning from graph structure and node/edge features. In this talk, we introduce GNNs for link prediction. We used GNNs to reconstruct contact networks for epidemic simulations. The result will be briefly presented.
Sang June Lee
Title : On zero-sum free sequences contained in random subsets of finite cyclic groups
Abstract : Let $C_n$ be a cyclic group of order $n$. A {\it sequence} $S$ of length $\ell$ over $C_n$ is a sequence $S = a_1\bdot a_2\bdot \ldots\bdot a_{\ell}$ of $\ell$ elements in $C_n$, where a repetition of elements is allowed and their order is disregarded. We say that $S$ is a zero-sum sequence if $\Sigma_{i=1}^{\ell} a_i = 0$ and that $S$ is a zero-sum free sequence if $S$ contains no zero-sum subsequence. In 2000, Gao obtained a construction of all zero-sum free sequences of length $n-1-k$ over $C_n$ for $0\leq k\leq \left\lfloor n/3\right\rfloor$. In this talk, we consider a generalization for a random subset of $C_n$. Let $R=R(C_n,p)$ be a random subset of $C_n$ obtained by choosing each element in $C_n$ independently with probability $p$. Let $N^R_{n-1-k}$ be the number of zero-sum free sequences of length $n-1-k$ in $R$. Also, let $N^R_{n-1-k,d}$ be the number of zero-sum free sequences of length $n-1-k$ having $d$ distinct elements in $R$. We obtain the expectation of $N^R_{n-1-k}$ and $N^R_{n-1-k,d}$ for $0\leq k\leq \left\lfloor n/3\right\rfloor$ and show that $N^R_{n-1-k}$ and $N^R_{n-1-k,d}$ are asymptotically almost surely (a.a.s.) around its expectation when $k$ is fixed. Moreover, we provide two ways to compute the expectations using partition numbers and a recurrence formula. This is joint work with Jun Seok Oh (Jeju University).
Jung Rye Park
Title: Introduction to topological data analysis using intuitive examples
Abstract : Topological Data Analysis (TDA) is an approach to the analysis of datasets using techniques from topology. The motivation is to study the shape of data. TDA is premised on the idea that the shape of data sets contains relevant information. However, computational complexity makes it impossible to quickly process many samples of higher dimensions. So, we will construct a preprocessing strategy, called the Characteristic Lattice Algorithm (CLA), that reduces the number of points to shorten the computation time for finding a persistent diagram, while maintaining the topological characteristics of the metric space. In this talk, we introduce TDA with a very simple example of a 2-dimensional array. And we applied TDA and CLA to the MNIST digits dataset, where the MNIST dataset is a large database of handwritten digits.