Hankyu is a third-year Ph.D. student in Computer Science at the University of Iowa specializing in Computational Epidemiology. He received his master’s degree in Data Science from Indiana University Bloomington in the year 2018.
Ph.D. in Computer Science
Hankyu is graduate research assistant of the Computational Epidemiology group at the University of Iowa.
One line of research is to propose non-pharmaceutical interventions to mitigate the disease spread. He has designed agent-based simulators to simulate disease spread in a unit level of a hospital. He built simulators for infectious diseases such as MRSA and COVID-19 and proposed low-cost interventions that can reduce the infection spread.
Another line of research is about identifying asymptomatic cases of a common hospital acquired infection called C. diff infections (CDI). He proposed a two-stage MLP model that (i) identifies asymptomatic cases that have missing labels, then (ii) evaluates their effect in the CDI spread.
He is interested in social network analysis in general, such as node classification, link predictions, and embeddings. Recently, he proposed a novel link prediction method in multi-relational networks; he looked into user communications in different channels on an online health community and utilized the heterogeneous information in these networks.
M.S. in Data Science
During his master’s degree, he gained knowledge of Artificial Intelligence. He did several research projects that apply Deep Learning, Machine Learning, and Reinforcement Learning techniques to solve problems.
In the field of Computer Vision, he built a CNN-RNN in an encoder-decoder scheme for Image Captioning from scratch using Keras with two other colleagues. For the encoder (Convolutional Neural Networks) he used VGG16, VGG19, and ResNet50 models to generate bottleneck features from images, using pre-trained weights. He used LSTM network as a decoder (Recurrent Neural Networks) to generate sentences using word embedding as input. He trained the LSTM and word embeddings to learn a mapping between image features and training captions.
In the area of Signal Processing, he and two other colleagues built a module that detects ambulance siren from traffic sounds and then locates the ambulance. He explored several Machine Learning and dimensionality reduction algorithms to solve this problem. He had an idea of giving ear to self-driving cars so that when the vehicle hears an ambulance coming, the module will help the vehicle to detect the ambulance and pull over.
He also has experience in Reinforcement Learning, a subfield in Artificial Intelligence. He explored different AI agents on pathfinding problems. As for a research project, he designed a novel environment “BusGridworld” that could be utilized to check the adaptiveness of the agents to the non-stationary world.
He enjoys solving real-world problems using AI approach. He applied Hierarchical Agglomerative Clustering to group 100 students into subgroups in a class based on their preferences. He loves playing chess; apart from playing chess, he built a chess agent using Minimax with Alpha-Beta pruning that could search through a few depths within seconds.
He is exploring various domains to apply the learning techniques in Artificial Intelligence: biomedical data, images, signal processing, social media, etc. He is dedicated to solving problems using his toolkits.