레퍼런스 흐름에 기반한 디자인 패턴의 확장 지점 식별

김희천,박찬진,김택수,유찬우,이형원,Kim, Hee-Chern,Park, Chan-Jin,Kim, Taek-Su,Yoo, Chan-Woo,Lee, Hyung-Won 한국정보처리학회 2012 정보처리학회논문지D Vol.19 No.4

디자인 패턴은 기존 소프트웨어의 설계 정보를 추상화하는 단위로 사용될 수 있다. 소프트웨어의 기능을 확장하기 위해 설계를 파악하고자 할 때 디자인 패턴의 인스턴스를 파악하는 것 만으로는 충분치 않을 수 있는데, 이는 기능의 확장이 일어나는 지점이 패턴 인스턴스 바깥에 존재할 수 있기 때문이다. 본 논문에서는 디자인 패턴의 핵심적인 구조를 구성하는 과정으로서 레퍼런스 흐름을 정의하였으며, 레퍼런스 흐름을 이용하여 패턴의 확장 지점을 찾아내는 방법을 제시하였다. Sometimes it is not sufficient to identify design pattern instances for extension of software, because the extension points exist outside of the instances. We define reference flows, which is a process of composing an intrinsic structure of design patterns, and suggest an analysis method based on reference flows for identifying the extension points.

GIS 기반 수산자원정보시스템의 개발 방법론에 관한 연구

박희서 ( Hee-seo Park ),김희천 ( Hee-chern Kim ) 한국정보처리학회 2012 한국정보처리학회 학술대회논문집 Vol.19 No.1

환경친화적 수산자원을 조성하고, 어족 자원 생산성을 극대화하기 위해 관련업계에서는 GIS 기반의 수산자원정보시스템을 구축하고 있다. 본 연구에서는 기존의 GIS 기반 정보시스템 개발방법론을 분석하고 수산자원정보시스템의 특성을 고려하여 기존 방법론에서 개선해야 할 사항들을 파악하였다. 연구의 결과로 GIS 데이터의 마이그레이션 작업을 위한 방법을 제시하였으며 실제 프로젝트에 적용하여 그 유용성을 확인하였다.

개인화 추천 시스템의 머신러닝을 위한 음수 미포함 행렬 분해와 변수 선택 기법의 비교

유찬우(Chan-Woo Yoo),김희천(Hee-Chern Kim) 한국디지털콘텐츠학회 2020 한국디지털콘텐츠학회논문지 Vol.21 No.4

NNMF(nonnegative matrix factorization) is widely used as a baseline machine learning model in personalized recommendation domain. Decomposed matrix by NNMF used to be applied to calculate missing elements in the original matrix before, recently they have been used as embeddings to search for similar items to a users favorite items by nearest neighbor method, because calculation of missing elements has to be done on all candidate items and it incurs delay in predictions. For nearest neighbor search, it is important to preserve cosine similarity between data points, so it needs to examine how NNMF preserves cosine similarity. In this paper, a comparison between NNMF and feature selection methods based on variance and entropy is performed in cosine similarity preservation perspective. The results shows feature selection methods are superior to NNMF when dimension is greater than some level. It is also shown that feature selection method is better than NNMF in time complexity.

JIRA 도구를 이용한 테스트 프로세스 관리

정학선(Hak-Sun Jeong),김희천(Hee-Chern Kim) 한국정보과학회 2015 한국정보과학회 학술발표논문집 Vol.2015 No.6

개인화 추천을 위한 비모형 차원 축소 기법

유찬우(Chan-Woo Yoo),김희천(Hee-Chern Kim) 한국디지털콘텐츠학회 2020 한국디지털콘텐츠학회논문지 Vol.21 No.3

Dimensionality Reduction methods are used to generate embeddings of users or items in machine learning models for personalized recommendation. Methods like SVD(singular valued decomposition) are hard to apply when big data is given in industrial fields because of their time and space complexity, although their mathematical properties are well known. In such a case, dimensionality reduction methods without models like random projection, which is fast and requires small space, are used. In this paper, another fast dimensionality reduction method without a model is proposed, which is better than Random Projection in cosine similarity preservation perspective. Experiments are performed with open datasets in personalized recommendation domain. The results shows the proposed method is better than random projection in cosine similarity preservation and execution time in all dimensions. A comparison between the proposed method and SVD, which creates a mkodel, is also performed. The proposed method appears that it achieves the similar degrees of cosine similarity preservation to SVD when the dimension is higher than some level and better execution time.

소규모 웹 응용을 위한 클라이언트 측 자바스크립트 프레임워크의 구축

김선종(Sun-Jong Kim),김희천(Hee-Chern Kim) 한국정보기술학회 2016 한국정보기술학회논문지 Vol.14 No.3

When increasing user requests for web service cause the lack of server resources, it is difficult for the server to provide services. However, if small IT companies install more server computers to solve this situation, they will face not only additional server costs but also license problem for using web application server. So we need a new effective method to handle more user requests on existing servers with restricted resources. In this paper, we suggest and implement a web-based application system architecture deploying a JavaScript framework on client-side browser which replaces the function of a web application server. We conduct experiments for performance evaluation to show the usefulness of our result compared with a general server-based counterpart. The experimental results show that our implementation can serve more concurrent users under the same server condition without speed loss from the user’s point of view.

래퍼를 이용한 웹 정보 추출 시스템

박유신(Yu-Shin Park),김희천(Hee-Chern Kim) 한국정보기술학회 2012 Proceedings of KIIT Conference Vol.2012 No.5

본 논문에서는 웹 페이지에 존재하는 정보를 추출하기 위한 래퍼 프로그램을 생성하기 위한 시스템을 제시한다. 웹 통합 검색과 다르게 웹에서 원하는 자료를 수집하기 위해서는 래퍼를 기반으로 하는 시스템이 필요하다. 스크립트 파일은 파이썬 언어를 이용하여 래퍼 기능을 구현한다. 그리고 스크립트 파일과 함께 사용될 추출 규칙이 정의된 엔진을 제공한다. 제안하는 시스템은 사용자가 이해하기 쉽게 정의된 규칙을 이용하여 스크립트 파일을 쉽게 작성할 수 있다. 또한 HTML 페이지에서 테이블 형태로 되어 있는 자료를 효율적으로 가져올 수 있는 방법을 제시한다. This paper presents a system to generate a wrapper program, which is for extracting existing information on a webpage. It requires a wrapper-based system to collect desired data from a webpage unlike an integrated search. A script file that is functioning as a wrapper realizes it by using a python programming language. It provides a defined extraction rule, which is to be used with a script file. The proposed system allows users to make a script easily using a defined rule, which is understandable for the users in an ease. It also shows a method fetching data efficiently that are in a table format on a HTML page.

GUI 기반 기능 테스트 자동화 스크립트의 재사용성 향상 기법

김성빈 ( Sung-vin Kim ),김희천 ( Hee-chern Kim ) 한국정보처리학회 2014 한국정보처리학회 학술대회논문집 Vol.21 No.1

기능 테스트 자동화를 위해서는 스크립트 구현에 많은 노력이 소요된다. 또 구현된 스크립트의 재사용성이 떨어져 소프트웨어 UI가 변경되면 수정이 필요하고 여러 어려움이 따른다. 본 연구의 목적은 GUI 기반 시스템 테스트 자동화의 특성과 문제점을 분석하여 테스트 스크립트의 재사용성을 향상시키는 방법을 제시하고 적용 사례를 보임으로써 테스트 자동화 작업의 생산성을 높이고 안정적이고 신뢰성이 있는 자동 테스트 결과를 확보하는데 있다.

컴퓨터 프로그래밍 교육을 위한 딥러닝 수업 설계

강환수(Hwan-Soo Kang),김희천(Hee-Chern Kim) 한국디지털콘텐츠학회 2020 한국디지털콘텐츠학회논문지 Vol.21 No.10

Artificial intelligence, a core technology in the era of the fourth industrial revolution, has recently become a basic learning competency for students to learn. Various educational studies for artificial intelligence are underway in the United States and the United Kingdom, and cases of artificial intelligence education research for elementary and secondary students are also beginning in Korea. In this study, a deep learning lecture was designed by deriving the learning competencies of computer programming education-centered classes for artificial intelligence education required for the 4th industrial revolution for computer major students. The deep learning lecture consisting of coding on Python, numpy, matplotlib and tensorflow, understanding artificial intelligence and deep learning, and implementing artificial and convolutional neural networks with tensorflow is applied to the actual class of D-university. In order to verify the effectiveness of learning competency of the class, a paired samples t-test was used. As a result of the analysis, there was significant differences in the pre- and post-test results of the learning competencies of deep learning proposed in this study.

파이썬 기반의 컴퓨터 프로그래밍 교육 모델 연구

강환수(Hwan-Soo Kang),이종만(Jong-Man Lee),김희천(Hee-Chern Kim) 한국디지털콘텐츠학회 2020 한국디지털콘텐츠학회논문지 Vol.21 No.4

Python is being chosen a lot as a programming language course to enhance computational thinking of non-computer majors in domestic universities. But for computer majors, Pythons choice is uncommon. In this study, to effectively educate computer major students about programming, we derived the learning competencies factors of Python programming for majors and presented a Python education model - CS² based on learning abilities. We applied this model to the actual class of python programming for D-University computer majors, and analyzed data before and after class with a paired samples t-test. As a result of the analysis, there were significant differences between the pre and post-application of the CS² education model. We hope that the educational model of this study will provide a turning point when Python is widely used in programming education for computer majors.