seungjin@postech.ac.kr

kshkawa@postech.ac.kr

My major research interest is to build an efficient machine learning algorithms for large-scale image search. Specifically, I'm interested in learning to hash for an efficient multimedia search system. Learning to hash is to estimate a hash function for a constant time complexity search. I'm trying to develop a hashing algorithm for multi-view or multi-modality dataset. In addition, I'm also interested in dealing with web-scale image dataset (more than 1 billion), and developed a practical near-duplicate image discovery system on a 1 billion web-image dataset. In order to deal with web-scale dataset, I'm trying to implement a various machine learning algorithm with MapReduce framework.

stonecold@postech.ac.kr

My primary research area is statistical machine learning with Bayesian nonparametric models. Currently, I'm interested in developing efficient posterior inference algorithms for Bayesian nonparametric models. I'm also interested in developing novel Bayesian nonparametric priors to solve complex real-world problems.

caster@postech.ac.kr

My research objective is modeling complex interactions in high dimensional large-scale time series from various domains. Main research interest is large-scale inference with probabilistic graphical model. Nowadays, I am trying to solve interaction related problems with various Machine Learning algorithms such as Topic Models, Neural Networks for timeseries and various point processes.

jtkim@postech.ac.kr

The research topic which I have been fascinated is automation of machine learning. Human machine learning experts have led the success of machine learning, applied in various cutting-edge technology industries, in recent years. However the tremendous attention from industries and also academia was not able to make non-experts apply off-the-shelf machine learning algorithms without domain knowledge, although approachable methods through open source packages have been developed. In this context, Bayesian optimization can be used to optimize various types of parameters. Therefore I think it will be a potent strategy for finding the extrema of objective functions that are expensive to evaluate.

yys8646@postech.ac.kr

wbkim91@postech.ac.kr

kimnay@postech.ac.kr

jiyuuyi@postech.ac.kr

My research topic is deep learning. Especially, I have interest in CNN architectures which facilitate training extremely deep networks by adopting the skip connections. These days, combining advantages of the skip connections with LSTM cells, I'm trying to improve performance of LSTM networks.

ihcho@postech.ac.kr

My research topic is mainly generative models. Especially, I am interested in Generative Adversarial Networks(GAN) among them. GAN is relevantly new compared to other generative models but it has received much attention because of convincing generated data and a new training paradigm. I am trying to solve unstable training issues and apply GAN to various applications.

minseop@postech.ac.kr

My research topic is mainly about deep learning for various image classification and algorithmic tasks. Especially, deep learning lacks a few shot generalization capacity due to the large size of parameters in the network and the way it learns from data. I'm working on that problem in the perspective of meta-learning and memory network with deep architecture.

einet89@postech.ac.kr

I am interested in hierarchical deep reinforcement learning, an area of research that aims to use deep neural networks as function approximators to identify and exploit structures in sequential decision problems. If successful, we will be able to create artificial agents capable of learning complex behaviors not even achievable by humans.

youngnk@postech.ac.kr