Joong-Ho (Johann) Won
Seoul National University · Computer Science
About the Lab
Professor Joong-Ho (Johann) Won's research lab specializes in statistical signal processing, machine learning, and high-performance computing, with a focus on developing robust estimation techniques for covariance matrices, advanced image and data analysis using wavelet-based models, and scalable algorithms for big data. The lab integrates statistical modeling with computational efficiency, particularly in medical imaging, dental arch analysis, and distributed computing environments like Hadoop and MPI. Key research directions include condition number-constrained estimation, shift-invariant multiresolution modeling, and parallelization of data-intensive algorithms in R and Python.
Research Overview
Research Output Trend
Figures are computed from collected data and may differ slightly.
Selected Papers
15In many signal processing applications, we want to estimate the covariance matrix of a multivariate Gaussian distribution. We often require the estimate to be not only invertible but also well-conditioned. We consider the maximum likelihood estimation of the covariance matrix with a constraint on the condition number. We show that this estimation problem can be reformulated as a convex univariate minimization problem, which admits an analytic solution. This estimation method requires no special
In block-based statistical texture segmentation approaches, modeling the global dependency between blocks as well as local statistics within a block is important for segmentation performance. A hidden Markov model (HMM) can be combined with a hidden Markov tree (HMT) to form an HMM-HMT model, which captures both global and local properties. Unfortunately, the real wavelet transform, on which the model is based, is not shift-invariant, which degrades the accuracy of the model. Further, its usual
We developed a novel visualization method for providing an uncluttered view of the abdominal aorta and its branches. The method abstracts the complex geometry of vessels using a convex primitive, and uses a sweep line algorithm to find a suboptimal placement of the primitive. The method was evaluated using 10 CT angiography datasets and resulted in a clear visualization with all cluttering intersections removed. The method can be used to convey clinical findings, including lumen patency and lesi
Orthodontists are interested in finding a set of standard arch forms for clinical orthodontic practice. In this paper, we propose a functional clustering method for the dental arches based on a mixture of U-shaped curves. We decide the number of clusters (equivalently, mixture components) using the Bayesian information criterion and the jump criterion based on a given distortion function. We apply our method to clustering the dental arch data from the nationwide standard occlusion study conducte
데이터의 규모가 급속히 증가하고 있는 현 시점에서 이러한 빅 데이터 처리를 위한 분산 컴퓨팅 환경으로 하둡(Hadoop)과 맵리듀스(Mapreduce)가 사실상의 표준으로 떠오르고 있으나, 이 환경에서의 실질적인 데이터 분석을 위한 방법론 개발에 대한 논의는 비교적 적은 편이다. 본 연구는 대표적인 네트워크 데이터 분석 알고리즘인 최대흐름 문제를 맵리듀스 분산 환경에서 처리하는 방법에 대해 소개하고, 맵리듀스를 기반으로 하는 대규모 최대흐름 알고리즘을 파이썬(Python) 언어를 이용하여 구현한 뒤, 하둡 환경에서 수행해보았다. 랜덤 네트워크와 영상 분할 데이터를 이용한 실험을 통해 수행 시간 및 확장성을 측정하였다. 랜덤 네트워크 실험에서는 정점의 수를 1000개부터 100만개까지 사용하였고, 영상 분할 실험에서는 정점의 수를 약 20개부터 크게는 약 17000개까지 사용하여 단일 서버와 분산 서버에서의 수행 시간을 비교하는 실험을 하였다. 이를 통해 맵리듀스 기반 알고리즘의 가능
We developed a novel visualization method for providing an uncluttered view of the abdominal aorta and its branches. The method abstracts the complex geometry of vessels using a convex primitive, and uses a sweep line algorithm to find a suboptimal placement of the primitive. The method was evaluated using 10 CT angiography datasets and resulted in a clear visualization with all cluttering intersections removed. The method can be used to convey clinical findings, including lumen patency and lesi
We study parallel processing techniques for the R programming language of high performance computing technology. In this study, we used massively parallel computing system which has 25,408 cpu cores. We conducted a performance evaluation of a distributed memory system using MPI and of a the shared memory system using OpenMP. Our findings are summarized as follows. First, For some particular algorithms, parallel processing is about 150 times faster than serial processing in R. Second, the distrib
로마자 서체에 대한 수치적 분류체계는 잘 발달되어 있지만, 한글 서체 분류를 위한 기준은 수치적으로 잘 정의되어 있지 않다. 본 연구의 목표는 한글 서체 분류를 위한 수치적 기준을 세우기 위해, 서체 스타일을 구분하는 중요한 특징들을 찾는 것이다. 컨볼루션 뉴럴 네트워크(convolutional neural network)를 사용하여 명조와 고딕 스타일을 구분하는 모형을 세우고, 학습된 필터를 분석해 두 스타일의 특징을 결정하는 피처(feature)를 찾고자 한다.
This study examines the characteristics of jade artifacts from The Niuheliang Site of the Transitional Phase in the Late Hongshan Culture, where most jade objects were discovered as grave goods. The analysis focuses on the functions of jade usage and the diffusion of jade-making techniques through comparative approaches. To explain the functional roles of jade artifacts and the spread of jade craftsmanship at Niuheliang, it is necessary to complement archaeological observation with a cultural an
초분광 영상 데이터는 픽셀마다 수백 개의 스펙트럼 밴드에 대한 정보가 주어지는 고차원 데이터로, 농업, 식품처리, 광물학, 물리학, 환경학, 지리학 등 광범위한 분야에 활용되고 있다. 그 중 하나는 토지 피복의 분류 문제인데, 이는 자연 재해 예방, 자연 자원 감시, 환경에 대한 정보 수집에 있어서 중요한 문제이다. 하지만 차원의 저주, 시공간적 변동성, 레이블된 데이터의 부족 때문에 토지 피복의 정확한 분류에는 어려움이 따른다. 이 논문에서는 이러한 문제를 해결하기 위해 컨볼루션 신경망에 기반한 새로운 심층 학습 구조를 제안한다. 제안된 구조는 원하는 지점 주변 픽셀의 정보를 컨볼루션 신경망을 통해 처리하고, 그 지점의 스펙트럼 정보를 강조하기 위해 컨볼루션 층의 출력과 스펙트럼 정보를 함께 소프트맥스 분류기의 입력으로 사용한다. 이 구조는 추가적인 특징 추출 과정을 필요로 하지 않고, 그래픽 처리 장치 등을 이용한 병렬화가 간편하다는 점에서 기존 방법들보다 유리하다. 실험 결과,
A network based on the modified probabilistic RAM architecture is proposed for VLSI implementation with on-chip learning capability. The pRAM structure is modified for the computation of error backpropagation and weight updates, where the multiplication and summations are performed with simple AND gates and OR gates, respectively. The simulation results for the derived algorithm are discussed, and a VLSI implementation of the pRAM for backpropagation is also explained.< <ETX xmlns:mml="http://ww
Good performance is important element not only in workplace but also in daily activities. Performance of the human depends on the mental capacity and mental workload. Especially, children in concrete operational stage is critical for further learning ability that they develop their ability to distinguish between quality and quantity. However, the reason that mental workload is difficult to quantify through physiological measures, makes it more complicated to demonstrate the mental workload. When
Research Areas
Dive deeper into Joong-Ho (Johann) Won's research on Nubint
Open this lab's papers in the app to read with AI, summarize, and cite in your writing.