Introduction to Neural Networks

윤충화 대한전자공학회 1989 대한전자공학회 단기강좌 Vol.- No.-

인터페이스 에이전트

윤충화,이형일,임윤택 한국정보처리학회 1998 정보처리학회지 Vol.5 No.1

이진 결정트리의 최적화기법에 관한 연구

김상귀,이형일,윤충화 明知大學校 産業技術硏究所 1997 産業技術硏究所論文集 Vol.16 No.-

Decision tree classifiers have been successfully used in many areas such as medical diagnosis, speech recognition, and expert systems, etc. The pattern classifier is basically an algorithm for partitioning the feature space into a number of disjoint regions. In many recognition problems for multi-featured patterns, nonparametric partitioning methods are usually employed to solve multi-featured and multi-class problems. In this study, we have proposed the AMIG-1 algorithm that can improve the recognition rate of the AMIG method. The AMIG algorithm uses the idea of average mutual information to partition the feature axes. We have used randomly generated real-valued data, Monk's data and Iris's data to validate the efficiency of the proposed algorithm.

인터페이스 에이전트에 관한 연구

윤충화,이형일,임윤택 明知大學校 産業技術硏究所 1998 産業技術硏究所論文集 Vol.17 No.-

Recently, many software developers have started to use the word 'Intelligent Software Agents' abusively, and a lot of research and development are under way in universities and research centers. Also the range of firms and universities actively pursuing agent technology is quite broad and the list is ever growing. Software agents have evolved from multi-agent systems (MAS), which in turn form one of the broad areas which fall under Distributed Artificial Intelligence (DAI). In this study, we have studied the various definitions and classifications concerning software agents, but the main emphasis is on the interface agents which act like personal assistants or teachers. We have studied characteristics of interface agents, and looked at various prototypes in detail, and analyzed their implementation methods.

지능형 검색 엔진을 위한 문서 분석 기법에 대한 연구

임윤택,윤충화 明知大學校 産業技術硏究所 2000 産業技術硏究所論文集 Vol.19 No.-

As we use machine learning algorithms for text categorization in information retrieval, the major difficulty arises when the number of features in a text pattern gets too large for ordinary machine learning algorithms to operate efficiently. In order to solve this problem, feature selection technique can be employed to select important features that affects the correct classification of texts, and in turn reduces the number of features in a text pattern. In this paper, we have compared Information Gain, Gini Index, Relief-F, and Document Frequency, which are major feature selection methods. Our experiment showed that by applying feature selection methods to reduce the number of features in a text pattern, we can actually achieve better classification performance than by simply constructing a pattern with all the unique words in a text.

EACH 시스템의 성능 향상 알고리즘

정태선,김상귀,윤충화 明知大學校 産業技術硏究所 1999 産業技術硏究所論文集 Vol.18 No.-

In this paper we proposed an improved algorithm that shows better performance than the orignal EACH system which is on the NGE (Nested Generalized Exemplar) theory. By allocating separate feature weights for each hyperrectangle and modifying the procedure for handling equi-distanced hyperrectangles, our algorithm is shown to be insensitive to the number of initial seeds and as a result, it is capable of incremental learning as more training patterns are available. We have experimentally verified that our algorithm shows better classification accuracy than the EACH system using the Iris and Monks benchmark data.

자생 참두메부추(Allium spirale Willd.)의 효과적인 육묘를 위한 생육환경조건

박충화,김정민,윤새미,임연진,윤미정,장보국 한국화훼학회 2023 화훼연구 Vol.31 No.3

참두메부추(Allium spirale Willd.)는 관상 및 식용식물로서 가치가 높다. 그러나, 참두메부추의 육묘를 위한 생육조건에 관한 연구는 부족한 실정이다. 본 연구는 육묘 플러그셀 크기, 차광율 및 시비처리가 참두메부추의 유묘 성장에미치는 영향을 조사하였다. 실험 결과, 플러그 셀 크기는 용적이 가장 큰 105셀에서 생육이 가장 우수하였다. 생육지표중 초장과 근장은 60% 차광처리에서 상대적으로 높게 측정되어 생육이 우수하였다. 속효성 고형비료[DO-Peters(N:P:K= 20:20:20)는 셀 당 100mg을 시비했을 때 근수와 근장의 수치가 가장 높게 조사되었으며, 속효성 액체비료 Peters(N:P:K= 20:20:20)는 8mL씩 주 2회 시비한 처리구에서 초장과 엽수의 수치가 가장 높았다. 따라서 참두메부추 유묘의 초기 생육에 있어서 속효성 고형비료는 100mg, 액체비료는 8mL를 주2회 시비하는 것이 적절하였다. 참두메부추는 원예상토가 충진된 105셀 플러그 트레이에 종자를 파종한 후 DO-PRO 100mg 또는 Peters 8mL를 주 2회 엽면시비하면서 60% 차광조건에서 재배하는 것이 효과적이었다. Allium spirale Willd. is valuable as it is an ornamental and edible plant. However, the research concerning the growth conditions for raising the seedlings of A. spirale is insufficient. In this study, the effects of seedling plug tray cell sizes, light shading rates, and fertilization treatments on the growth of seedlings were investigated. The growth was the best in the 105-cell plug tray, which had the largest tray volume. Of the growth indicators tested, plant growth and root length were relatively high under 60% shade. Applying 100 mg of the fast-acting coated fertilizer [DO-PRO (N:P:K = 12:8:12 + 6CaO)] per cell resulted in the highest root number and length values. The short-acting liquid fertilizer [Peters (N:P:K = 20:20:20)], when applied at 8 mL twice a week rather than once a week, resulted in an enhanced plant length and increased number of leaves. Therefore, applying 100 mg of the fast-acting solid fertilizer and 8 mL of the short-acting liquid fertilizer per cell twice a week was considered appropriate for the initial growth of seedlings. Thus, the most effective growth conditions were sowing in a 105-cell plug tray filled with horticultural substrate, addition of 100 mg of DO-PRO or 8 mL of Peters twice a week, and cultivation under 60% shading.

고정 분할 평균기법을 사용하는 향상된 메모리 기반 추론

정태선,이형일,윤충화 明知大學校 産業技術硏究所 1999 産業技術硏究所論文集 Vol.18 No.-

In this paper, we have proposed the FPA(Fixed Partition Averaging) algorithm in order to improve the storage requirement and classification time of Memory Based Reasoning method. The proposed method enables us to use the storage more efficiently by extracting representatives out of training patterns. After partitioning the pattern space into a fixed number of equally-sized hyperrectangles, it averages patterns in each hyperrectangle to extract a representative. Also we have used the mutual information between the features and classes as weight for features to improve the classification performance.

자생 부추속(Allium L.) 식물 3종의 차광 정도에 따른 생육 변화

박충화(Chung Hwa Park),윤새미(Sae Mi Yoon),김정민(Jeong Min Kim),김영재(Young Jae Kim),김상용(Sang Yong Kim),정지영(Ji Young Jung) 한국원예학회 2020 한국원예학회 학술발표요지 Vol.2020 No.11

플러그 트레이 셀의 크기, 차광률, 시비조건이 강부추와 갯부추의 육묘에 미치는 영향

박충화,장보국,김정민,최혁재,윤미정 한국화훼학회 2023 화훼연구 Vol.31 No.4

Among the native Allium plants, A. thunbergii for. rheophytum ined. and A. pseudojaponicum Makino were valuable as ornamental, edible, and medicinal resources, but they need to be studied due to insufficient investigation of growth environmental conditions for seedling. This study was conducted to investigate the effect of plug tray cell size, light shading rates, and fertilization treatments on the seedling of two allium species. As a result of cultivating in the plug cell size, 50 cells with the largest volume among 50, 72, 105, 128, 162, and 200 cell treatments showed excellent growth of plant length, number of leaves, number of roots and root length. However, it is considered effective to use a 128-cell plug tray in terms of production cost and efficiency during seedling. The seedlings according to the light shading rates were significantly higher plant length, number of roots, and root length in 30~60% shade. When the growth indicators except for fresh and dry weight were investigated in the fertilizer treatment, A. thunbergii for. rheophytum were 0.1 g of fast-acting solid fertilizer (DO-PRO) and A. pseudojaponicum were 8 mL liquid fertilizer treatment once a week, and both species were effective in fast-acting fertilizer treatment. The most effective growth conditions for two allium species seedlings were found to be sowing in a 128-cell plug tray filled with horticultural substrate, fertilization of 0.1 g of DO-PRO or 8 mL of Peters once a week, and cultivation under 30~60% shading treatment.