Research

The problem of finding the information source is to identify the true source of information spread. This is clearly of practical importance, because harmful diffusion can be mitigated or even blocked, e.g., by vaccinating humans or installing security updates. We study intensively on the following approaches.

• Finding Rumor Source Using Side Information from Querying and Protectors 

How can we infer the rumor sources using random snapshots, using side information from querying and protectors? What are the required querying budgets for arbitrary detection efficiency? 

• Estimating the Information Source under Decaying Diffusion Rates 

• Designing an Efficient Algorithm for Inferring the Information Source under Dynamic Networks

Due to a rapid development of the Internet, huge information flows over the network. In such a situation, if a virus or malicious information is received in a place such as vehicular network, it can cause a dangerous accident. To avoid this, we study on designing efficient algorithms to detect the abnormal information (or source node) quickly using online learning techniques. 

• Abnormal Information Detection

• Abnormal Node (or Source) Detection

The more complete development of autonomous vehicles is one of the key technologies driving the fourth industrial revolution. This is due to the fusion of software technologies such as communication and AI as well as classical dynamics and control. We study on the platooning driving based on Reinforcement Learning (RL), in which various vehicles on the road communicate with each other while learning information about the environment with one agent in relation to autonomous driving.

• Distance Learning using Communication 

• Learning the Cluster Header Selection under Dynamic Platooning

Recently, there have been many studies about the spreading phenomena of information/product due to developing of online Social Network Service over social networks such as Facebook and Twitter, etc. Especially, some studies consider two competitive information and/or product diffusions by a game model which called seeding game. In this game, each player chooses a seed set who spreads their information initially in the network that maximizes the diffusion at the end of diffusion process. These are modeled by non-adaptive manner in the sense that the player chooses all possible seed nodes before the diffusion.  Different to this, we are interested in adaptive seeding game for information diffusion that each player can choose their seed nodes after partial observation of the diffusion snapshot to maximizes the number of diffused nodes for their own information. 

• Competitive Seeding Game 

• Cooperative Seeding Game 

Among the various variables that determine the performance of next-generation semiconductors, charge trap is difficult to analyze quantitatively, making it difficult to clearly identify the trigger. In semiconductor materials and devices, charge traps are a factor that causes charge carriers to be trapped inside or outside of the material, interfering with charge behavior and significantly affecting performance degradation. However, a clear principle that causes this has not yet been revealed. In particular, changes in the performance of transistors are determined by complex interactions due to various variables such as the type and crystal structure of the semiconductor, the optical properties, thickness and roughness of the semiconductor, the dielectric constant of the insulating film, and the electrode junction structure. Research on charge trap changes due to these individual variables is difficult to control delicate variables, and quantitative analysis is insufficient. Recently, a variety of machine learning, such as deep learning, has been applied to predict and analyze semiconductor performance. However, charge traps are an uncontrollable factor due to the absence of appropriate measurement technology and must be identified, but the lack of explanation of the causal relationship of data is an obstacle to semiconductor development.

• Identification of Relationship of Semiconductor Charge Traps based on Probabilistic Causal Model 

• XPS analysis based on statistical Bayes' theorem and intervention technique and development of a causal model for electrical properties of oxygen vacancy in oxide semiconductors