http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
PERFORMANCE ENHANCEMENT OF PARALLEL SYMMETRIC GENERALIZED EIGENVALUE SOLVER
Si Hyong PARK,Jong Geun MOON,Wan Il BYUN,Seung Jo KIM 한국산업응용수학회 2005 한국산업응용수학회 학술대회 논문집 Vol.- No.-
We discuss the performance enhancement techniques for a parallel block Lanczos eigen-solver which deals with a symmetric generalized eigenvalue problem. Approaches proposed in this research are suited for distributed memory parallel environments. Performance tuning of a eigen-solver covers the efficient implementation of block Lanczos iteration and the optimization of the linear equation solver incorporated with the eigen-solver. We propose a block Lanczos iteration equipped with effective mass multiplication algorithm and apply some kinds of concepts to enhance the performance of a linear equation solver. Factorization and triangular system solving phases are simultaneously considered for optimization of a linear equation solver. After individual routines are optimized respectively, they are applied to a single eigen-solver code.
DESIGN OF PARALLEL BLOCK LANCZOS CODE BASED ON DATE STRUCTURE OF MULTIFRONTAL SOLVER
Si Hyong Park,Ji Joong Moon,Seung Jo Kim 한국산업응용수학회 2005 한국산업응용수학회 학술대회 논문집 Vol.- No.-
We discuss the assigning scheme of the Lanczos vectors for distributed-memory parallel architectures with the MPI protocol. The eigenvalue problem investigated belongs to the symmetric generalized one with the mass matrix which results from the finite element method. The distributing idea of the Lanczos vectors is developed from the data structure of the multi frontal linear solver. Each processor stores only its own part of the mass matrix and the Lanczos vectors, and the inner product about the mass matrix is simply computed by one call of collective communication about a scalar. In addition, the solution procedure for a linear equation originated from the generalized eigenvalue problem is started with the Lanczos vectors without modifications or communications. Finally, the performance of the code developed is enhanced by using the level 3 BLAS which can be applied only to the block Lanczos algorithm. Using the code developed with such an idea, the performance and the scalability are tested and compared with commercial codes.
Clinicopathological Significance of Elevated PIK3CA Expression in Gastric Cancer
Jang, Si-Hyong,Kim, Kyung-Ju,Oh, Mee-Hye,Lee, Ji-Hye,Lee, Hyun Ju,Cho, Hyun Deuk,Han, Sun Wook,Son, Myoung Won,Lee, Moon Soo The Korean Gastric Cancer Association 2016 Journal of gastric cancer Vol.16 No.2
Purpose: PIK3CA is often mutated in a variety of malignancies, including colon, gastric, ovary, breast, and brain tumors. We investigated PIK3CA expression in gastric cancer and explored the relationships between the PIK3CA expression level and clinicopathological features as well as survival of the patients. Materials and Methods: We examined PIK3CA expression in a tissue microarray of 178 gastric adenocarcinomas by immunohistochemistry and reviewed patients' medical records. Results: In our study, 112 of the 178 gastric cancer patients displayed positive PIK3CA expression. Overexpression of PIK3CA was correlated with low grade differentiation (P=0.001), frequent lymphatic invasion (P=0.032), and high T stage (P=0.040). Patients with positive PIK3CA staining were more likely to display worse overall survival rate than those with negative PIK3CA staining, as determined by Kaplan-Meier survival analysis with log-rank test (P=0.047) and a univariate analysis using the Cox proportional hazard model (hazard ratio=1.832, P=0.051). Conclusions: Elevated PIK3CA expression was significantly correlated with tumor invasiveness, tumor phenotypes, and poor patient survival.
User modeling based on fuzzy category and interest for web usage mining
Si-Hun Lee,Jee-Hyong Lee 한국지능시스템학회 2005 INTERNATIONAL JOURNAL of FUZZY LOGIC and INTELLIGE Vol.5 No.1
Web usage mining is a research field for searching potentially useful and valuable information from web log file. Web log file is a simple list of pages that users refer. Therefore, it is not easy to analyze user's current interest field from web log file. This paper presents web usage mining method for finding users' current interest based on fuzzy categories. We consider not only how many times a user visits pages but also when he visits. We describe a user's current interest with a fuzzy interest degree to categories. Based on fuzzy categories and fuzzy interest degrees, we also propose a method to cluster users according to their interests for user modeling. For user clustering, we define a category vector space. Experiments show that our method properly reflects the time factor of users' web visiting as well as the users' visit number.
웹 사용 마이닝을 위한 퍼지 카테고리 기반의 프랜잭션 분석 기법
이시헌(Si-Heon Lee),이지형(Jee-Hyong Lee) 한국지능시스템학회 2004 한국지능시스템학회 학술발표 논문집 Vol.14 No.1
웹 사용 마이닝(Web usage mining)은 웹 로그 파일(web log file)이나 웹 사용 데이타(Web usage data)에서 의미 있는 정보를 찾아내는 연구 분야이다. 웹 사용 마이닝에서 일반적으로 많이 사용하는 웹 로그 파일은 사용자들이 참조한 페이지의 단순한 리스트들이다. 따라서 단순히 웹 로그 파일만을 이용하는 방법만으로는 사용자가 참조했던 페이지의 내용을 반영하여 분석하는데에는 한계가 있다. 이러한 점을 개선하고자 본 논문에서는 페이지 위주가 아닌 웹 페이지가 포함하고 있는 내용(아이템)을 고려하는 새로운 퍼지 카테고리 기반의 웹 사용 마이닝 기법을 제시한다. 또한 사용자를 잘 파악하기 위해서 시간에 따라 관심의 변화를 파악하는 방법을 제시한다.
User modeling based on fuzzy category and interest for web usage mining
Lee, Si-Hun,Lee, Jee-Hyong Korean Institute of Intelligent Systems 2005 INTERNATIONAL JOURNAL of FUZZY LOGIC and INTELLIGE Vol.5 No.1
Web usage mining is a research field for searching potentially useful and valuable information from web log file. Web log file is a simple list of pages that users refer. Therefore, it is not easy to analyze user's current interest field from web log file. This paper presents web usage mining method for finding users' current interest based on fuzzy categories. We consider not only how many times a user visits pages but also when he visits. We describe a user's current interest with a fuzzy interest degree to categories. Based on fuzzy categories and fuzzy interest degrees, we also propose a method to cluster users according to their interests for user modeling. For user clustering, we define a category vector space. Experiments show that our method properly reflects the time factor of users' web visiting as well as the users' visit number.