• Annual Conference on Human and Language Technology
• /
• 2019.10a
• /
• pp.8-11
• /
• 2019

앙상블 기법은 여러 모델을 종합하여 최종 판단을 산출하는 기계 학습 기법으로서 딥러닝 모델의 성능 향상을 보장한다. 하지만 대부분의 기법은 앙상블만을 위한 추가적인 모델 또는 별도의 연산을 요구한다. 이에 우리는 앙상블 기법을 교차 검증 방법과 결합하여 앙상블 연산을 위한 비용을 줄이며 일반화 성능을 높이는 교차 검증 앙상블 기법을 제안한다. 본 기법의 효과를 입증하기 위해 MRPC, RTE 데이터셋과 BiLSTM, CNN, BERT 모델을 이용하여 기존 앙상블 기법보다 향상된 성능을 보인다. 추가로 교차 검증에서 비롯한 일반화 원리와 교차 검증 변수에 따른 성능 변화에 대하여 논의한다.

• Korean Chemical Engineering Research
• /
• v.59 no.4
• /
• pp.532-541
• /
• 2021

Recently, research on the application of artificial intelligence in the chemical process has been increasing rapidly. However, overfitting is a significant problem that prevents the model from being generalized well to predict unseen data on test data, as well as observed training data. Cross validation is one of the ways to solve the overfitting problem. In this study, the time-series cross validation method was applied to optimize the number of batch and epoch in the hyperparameters of the prediction model for the 2,3-BDO distillation process, and it compared with K-fold cross validation generally used. As a result, the RMSE of the model with time-series cross validation was lower by 9.06%, and the MAPE was higher by 0.61% than the model with K-fold cross validation. Also, the calculation time was 198.29 sec less than the K-fold cross validation method.

• The Journal of the Korea Contents Association
• /
• v.18 no.9
• /
• pp.614-622
• /
• 2018

As multiple weapons are fired simultaneously in multi-target and multi-weapon environments, a possibility always exists in the collision occurred by the intersection between weapon trajectories. The collision between weapons not only hinders the rapid reaction but also causes the loss of the asset of weapons of friendly force to weaken the responsive power against the threat by an enemy. In this paper, we propose an intersection validation and interference elimination algorithm between weapon trajectories in multi-target and multi-weapon environments. The core points of our algorithm are to confirm the possible interference through the analysis on the intersections between weapon trajectories and to eliminate the mutual interference. To show the superiority of our algorithm, we implement the evaluation and verification of performances through the simulation and visualization of our algorithm. Our experimental results show that the proposed algorithm performs effectively the interference elimination regardless of the number of targets and weapon groups by showing that no cross point exists.

• Annual Conference on Human and Language Technology
• /
• 2020.10a
• /
• pp.103-107
• /
• 2020

의존구조 말뭉치는 자연언어처리 분야에서 문장의 의존관계를 파악하는데 널리 사용된다. 이러한 말뭉치는 일반적으로 오류가 없다고 가정하지만, 현실적으로는 다양한 오류를 포함하고 있다. 이러한 오류들은 성능 저하의 요인이 된다. 이러한 문제를 완화하려고 본 논문에서는 XGBoost와 교차검증을 이용하여 이미 구축된 구문분석 말뭉치로부터 오류를 탐지하는 방법을 제안한다. 그러나 오류가 부착된 학습말뭉치가 존재하지 않으므로, 일반적인 분류기로서 오류를 검출할 수 없다. 본 논문에서는 분류기의 결과를 분석하여 오류를 검출하는 방법을 제안한다. 성능을 분석하려고 표본집단과 모집단의 오류 분포의 차이를 분석하였고 표본집단과 모집단의 오류 분포의 차이가 거의 없는 것으로 보아 제안된 방법이 타당함을 알 수 있었다. 앞으로 의미역 부착 말뭉치에 적용할 계획이다.

• Journal of Korean Society of Transportation
• /
• v.20 no.4
• /
• pp.109-122
• /
• 2002

본 연구에서는 교차로의 비용 및 특성을 고려한 도로선형최적화 모형을 유전자 알고리즘(Genetic Algorithms)을 이용하여 개발하였다. 기존의 도로선형최적화 모형은 교차로 특성을 고려하지 못해서 실제 적용에 심대한 문제점을 내재하고 있다. 본 논문에서는 특정 도로선형에 교차로 건설의 필요가 있을 경우, 민감(Sensitive)하고 지배적인(Dominating) 교차로 비용 항목들 즉, 토공비용, 보상비, 포장비, 사고비용, 지체 및 연료소모비용 등의 산정이 시도되었다. 또한 비교적 우수한 도로선형 대안을 유전자 알고리즘을 이용한 탐색과정 중에서 비효율적으로 강제 퇴화시키는 단점 보완을 위한 교차로 국소 최적화 방법(Local Optimization of Intersections)이 개발되어 기존 모형을 보완하였다. 공간상의 도로선형은 매개변수적 묘사(Parametric Representation)를 통하여 구현하였으며 벡터운영(Vector Manipulation)을 통해 교차로비용 산정의 근간인 교차점과 다른 중요점들의 좌표를 찾을 수 있었다. 개발된 교차로 비용산정 모형이 보다 정밀하게 교차로 비용을 산정함이 증명되었으며 궁극적으로는 기존의 최적화 모형의 단점을 보완할 수 있음이 제시되었다. 또한, 새로이 제시된 교차로 국소 최적화 방법이 최적대안 탐색과정의 유연성을 증대하였으며, 결과적으로 효율적인 교차로의 유지에 기여함을 알 수 있었다. 제시된 교차로 국소 최적화 방법은 추후 단일노선이 아닌 도로망 최적화시의 기초를 제시함은 주목할 만 하다. 두개의 예제에서 도출된 최적노선 및 교차로 비용 등의 검토 결과, 도로상의 교차로 건설비용은 도로선형 최적화에 큰 영향을 미치는 실질적이며 민감한 비용 항목임이 검증되었으며 이는 도로선형최적화 모형이 교차로 비용을 반드시 검토 및 평가할 수 있어야 함을 반증한다.

• Journal of Cadastre & Land InformatiX
• /
• v.50 no.1
• /
• pp.75-85
• /
• 2020

This study examines machine learning cross-validation to utilized create ROI for classification of land cover. The study area located in Sejong and one KOMPSAT-3A image was used in this analysis: procedure on October 28, 2019. We used four bands(Red, Green, Blue, Near infra-red) for learning cross validation process. In this study, we used K-fold method in cross validation and used SVM kernel type with cross validation result. In addition, we used 4 kernels of SVM(Linear, Polynomial, RBF, Sigmoid) for supervised classification land cover map using extracted ROI. During the cross validation process, 1,813 data extracted from 3,500 data, and the most of the building, road and grass class data were removed about 60% during cross validation process. Based on this, the supervised SVM linear technique showed the highest classification accuracy of 91.77% compared to other kernel methods. The grass' producer accuracy showed 79.43% and identified a large mis-classification in forests. Depending on the results of the study, extraction ROI using cross validation may be effective in forest, water and agriculture areas, but it is deemed necessary to improve the distinction of built-up, grass and bare-soil area.

• Korean Journal of Cognitive Science
• /
• v.10 no.1
• /
• pp.25-43
• /
• 1999

The focus of this study is on the elements of structure design that facilitate u user interaction with applications within cyberspace Structure design entails decisions regarding the optimal classification and hierarchical organization of information into s successively higher units. i.e .. the grouping of highly related information in the form of nodes of a site and the subsequent connection of nodes that are inter-related. The decisions are based on the designer's subjective classification framework. which is not always compatible with that of the user. We propose that the ensuing cognitive dissonance can be reduced via the employment of multiple hierarchies and cross-referencing links. Multiple hierarchies represent a single information space in terms of a number of single hierarchies. each of which represent a different perspective Cross-referencing refers to the inter-connection between the constituent hierarchies by providing a link to the alternate hierarchy for information that is most likely to be categorized in diverse manners by users with differing perspectives. In this study we conducted two empirical studies to gauge the effectiveness of multiple hierarchies and Cross-referencing links in the domain of cyber shopping malls. In the first phase. an experiment was conducted to determine how subjects classified given products with respect to two different perspectives for categorization. Experimental cyber malls were developed based on the results from the first phase to test the effectiveness of multiple hierarchies and cross-referencing links. Results show that the ease of navigation was higher for cyber malls that had implemented cross-referencing links are of greater value when used in conjunction with single hierarchical designs rather than multiple hierarchies. Users satisfaction with and ease of navigation was higher for cyber malls that had not implemented multiple hierarchies. This paper concludes with discussion of these results and their implications for designers of cyber malls.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.22 no.6
• /
• pp.1-16
• /
• 2023

Accurate traffic information prediction is considered to be one of the most important aspects of intelligent transport systems(ITS), as it can be used to guide users of transportation facilities to avoid congested routes. Various deep learning models have been developed for accurate traffic prediction. Recently, ensemble techniques have been utilized to combine the strengths and weaknesses of various models in various ways to improve prediction accuracy and stability. Therefore, in this study, we developed and evaluated a traffic information prediction model using various deep learning models, and evaluated the performance of the developed deep learning models as a stacking ensemble. The individual models showed error rates within 10% for traffic volume prediction and 3% for speed prediction. The ensemble model showed higher accuracy compared to other models when no cross-validation was performed, and when cross-validation was performed, it showed a uniform error rate in long-term forecasting.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.34 no.5
• /
• pp.541-547
• /
• 2010

The rigorous validation of the accuracy of metamodels is an important topic in research on metamodel techniques. Although a leave-k-out cross-validation technique involves a considerably high computational cost, it cannot be used to measure the fidelity of metamodels. Recently, the mean$_0$ validation technique has been proposed to quantitatively determine the accuracy of metamodels. However, the use of mean$_0$ validation criterion may lead to premature termination of a sampling process even if the kriging model is inaccurate. In this study, we propose a new validation technique based on the mean and variance of the response evaluated when sequential sampling method, such as maximum entropy sampling, is used. The proposed validation technique is more efficient and accurate than the leave-k-out cross-validation technique, because instead of performing numerical integration, the kriging model is explicitly integrated to accurately evaluate the mean and variance of the response evaluated. The error in the proposed validation technique resembles a root mean squared error, thus it can be used to determine a stop criterion for sequential sampling of metamodels.

• The Korean Journal of Applied Statistics
• /
• v.14 no.2
• /
• pp.415-427
• /
• 2001

공간통계학에서는 다변량 공간자료에 대한 예측방법으로서 코크리깅 기법을 이용한다. 본 논문에서는 코크리깅을 위한 첫 번째 단계인 교차베리오그램의 추정에 대한 감도분석 대신에 일반통계학적 측면에서 주성분점수를 이용한 감도분석방법을 제안한다. 변수가 2개인 경우, 교차베리오그램에 대한 감조분석의 결과와 제안된 주성분점수를 이용한 감도분석의 결과를 비교해 본다. 모의실험을 통하여 제안한 방법의 타당을 검증하고, 실제 자료를 이용한 사례분석의 결과로써 재확인해 본다.