Machine Learning Group

Research Area

We develop algorithms and theories for statistical machine learning and graphical models, with applications to search and mining, biomedical data analysis and bioinformatics, and user interface. Of particular methods that we are working on are Bayesian learning, convex optimization, dimensionality reduction, kernel machines, manifold learning, matrix and tensor factorizations, probabilistic graphical models, semi-supervised learning, transfer learning.

Learning to Hash

Hashing refers to methods for embedding high-dimensional data into a similarity-preserving low-dimensional Hamming space such that similar objects are indexed by binary codes whose Hamming distances are small. Learning hash functions from data has recently been recognized as a promising approach to approximate nearest neighbor search for high-dimensional ... more

Multi-label Learning

Multi-label learning is an extension of multiclass learning. It seeks a classification function that predicts a set of relevant labels for an instance, whereas in multiclass learning a single label is assigned to an instance. Multi-label problems arise in various ... more

Multi-view Learning

Real-world data including image, documents may have multiple representation, and each of the representations may be called a 'view'. Multi-view learning refers to the methods that utilizes multiple representations of data. In multi-view learning ... more


This novel assitive technology recognizes the direction of the tongue movement to provide an alternative communication channel to help the persons in daily tasks, especially who has limb motor disabilities. ... more

Semi-supervised Learning

In machine learning, a learner requires previous observations, called 'labeled data'. Each observation consists of (1) a set of feature values and (2) a target value what we want to predict, also called 'label'. Typically, feature ... more


Bioinformatics and computational biology involve the use or development of techniques including applied mathematics, informatics, statistics, computer science, artificial intelligence, chemistry, and biochemistry to solve ... more

Brain Computer Interface

Brain computer interface (BCI) is a system that is designed to translate a subject's intention or mind into a control signal for various devices. Electroencephalogram (EEG), which captures electrical potentials ... more

Search and Recommendation

Search and recommendation is to retrieve relevant information from large-scale databases. Our research is focused on personalized web search and movie recommendation, which are closely related to multi-relational networks ... more

Human-computer interaction

Human-computer interaction is the study of interaction between people and computers. A basic goal of HCI is to improve the interactions between users and computers by making computers more usable and receptive to the user's ... more

Nonparametric Method

Nonparametric methods are very flexible models for inferring non-linear functions, or solving density estimation problems. Specially, in the Bayesian approach, Gaussian processes and Dirichlet processes ... more

ICA/Source Separation

Independent component analysis (ICA) aims to extract independent components from their linear mixtures, without the aid of prior knowledge. Its important application is source separation which finds a set of original sources ... more

Matrix/Tensor factorization

Component analysis (e.g. PCA, ICA, FLDA, NMF) have been successfully applied in numerous visual, graphics and signal processing tasks over the last two decades. Recently, these methods are generalized to multilinear methods ... more

Research proposals for undergraduate students (written in Korean)