• Annual Conference on Human and Language Technology
• /
• 2014.10a
• /
• pp.56-60
• /
• 2014

기계학습 방법에 기반한 자연어 분석은 학습 데이터가 필요하다. 학습 데이터가 구축된 소스 도메인이 아닌 다른 도메인에 적용할 경우 한국어 의미역 인식 기술은 15% 정도 성능 하락이 발생한다. 본 논문은 이러한 다른 도메인에 적용시 발생하는 성능 하락 현상을 극복하기 위해서 기존의 소스 도메인 학습 데이터를 활용하여, 소규모의 타겟 도메인 학습 데이터 구축만으로도 성능 하락을 최소화하기 위해 한국어 의미역 인식 기술에 prior 모델을 제안하며 기존의 도메인 적응 알고리즘과 비교 실험하였다. 추가적으로 학습 데이터에 사용되는 자질 중에서, 형태소 태그와 구문 태그의 자질 값을 기존보다 단순하게 적용하여 성능의 변화를 실험하였다.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2018.01a
• /
• pp.243-246
• /
• 2018

본 논문에서는 2차원 공간상에서 상부에서 하부로 떨어지는 블록을 하부에서 피하는 게임에서 강화 학습에 사용되는 DQN 알고리즘을 이용하여 블록 피하기 에이전트를 구현하고 학습 통해 점점 더 높은 점수를 받는 모습을 확인하였다. 파이썬을 이용하여 게임을 구현한 후 텐서플로우를 이용하여 DQN를 이용한 에이전트를 구현하였다. 에이전트는 보상을 통한 학습을 통하여 점점 강화되도록 하였는데, 초기에는 무작위로 움직였으나, 환경으로부터 받는 보상으로 점점 더 능숙하게 피하는 모습을 관찰할 수 있었다. 본 구현에서는 4000번 정도의 게임 시행에서 아주 능숙하게 피하는 결과를 얻을 수 있었다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2010.06b
• /
• pp.184-187
• /
• 2010

본 논문에서는 터치 제스처의 인식을 위해 시도된 여러 방법 중 테이블탑 디스플레이상에서 HMM을 이용한 제스처의 학습과 사용에 대해 제안한다. 터치 제스처는 제스처의 획(stroke)에 따라 single stroke와 multi stroke로 분류할 수 있다. 그러므로 제스처의 입력은 영상프레임에서 터치 궤적에 따라 변하는 방향 벡터를 이용하여 방향코드로 분석될 수 있다. 그리고 분석된 방향코드를 기계학습을 통하여 학습시킨 후, 인식실험에 사용한다. 제스처 인식 학습에는 총 10개의 제스처에 대하여 100개 방향코드 데이터를 이용하였다. 형태를 갖추고 있는 제스처는 미리 정의되어 있는 제스처와 비교를 통하여 인식할 수 있다. (4 방향 드래그, 원, 삼각형, ㄱ ㄴ 모양 >, < ) 미리 정의되어 있는 제스처가 아닌 경우에는 기계학습을 통하여 사용자가 의미를 부여한 후 제스처를 정의하여 원하는 제스처를 선택적으로 사용할 수 있다. 본 논문에서는 테이블탑 디스플레이 환경에서 사용자의 터치제스처를 인식하는 시스템을 구현하였다. 앞으로 테이블탑 디스플레이 환경에서 터치 제스처 인식에 적합한 알고리즘을 찾고 멀티터치 제스처를 인식하는 연구도 이루어져야 할 것이다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2004.10b
• /
• pp.706-708
• /
• 2004

대부분의 기계학습 알고리즘은 학습 데이터에서 각각의 범주간의 비율이 동일하거나 비슷하다는 가정 하에 문제를 풀게 된다. 그러나 실제 문제에서는 그 비율이 동일하지 않으며 매우 큰 차이를 보이기도 하는데, 이는 분류 성능을 저하시키는 요인이기도 하다 따라서 본 논문에서는 이러한 데이터의 불균형 문제를 해소하는 방안으로 SVM 앙상블 기법을 적용한 샘플링을 제안하고 이를 실제 불균형 데이터에 적용함으로써 제안된 방법이 기존의 방법들에 비해 향상된 성능을 나타내는 것을 보였다.

• Journal of the Korea Society of Computer and Information
• /
• v.25 no.9
• /
• pp.193-201
• /
• 2020

In this study, we developed a simulation app that performs the Test-and-Set algorithm. The test-and-set algorithm is a highly difficult algorithm, so this simulation app was created for learners who have difficulty understanding it. Learners who want to understand the Test-and-Set algorithm gather to form a team, and use this simulation app to discuss and practice, and these teams can practice at the same time. The test-and-set, which is assumed to be a machine language, is not interrupted by using a queue, and it can be seen that the configured simulation app performs well in all three conditions of 'mutual exclusion', 'progress', and 'bounded waiting' that must be solved in the critical area problem.

• Korean Journal of Remote Sensing
• /
• v.34 no.5
• /
• pp.811-827
• /
• 2018

The purpose of this study is to compare machine learning algorithm and deep learning algorithm in crop classification using multi-temporal remote sensing data. For this, impacts of machine learning and deep learning algorithms on (a) hyper-parameter and (2) training sample size were compared and analyzed for Haenam-gun, Korea and Illinois State, USA. In the comparison experiment, support vector machine (SVM) was applied as machine learning algorithm and convolutional neural network (CNN) was applied as deep learning algorithm. In particular, 2D-CNN considering 2-dimensional spatial information and 3D-CNN with extended time dimension from 2D-CNN were applied as CNN. As a result of the experiment, it was found that the hyper-parameter values of CNN, considering various hyper-parameter, defined in the two study areas were similar compared with SVM. Based on this result, although it takes much time to optimize the model in CNN, it is considered that it is possible to apply transfer learning that can extend optimized CNN model to other regions. Then, in the experiment results with various training sample size, the impact of that on CNN was larger than SVM. In particular, this impact was exaggerated in Illinois State with heterogeneous spatial patterns. In addition, the lowest classification performance of 3D-CNN was presented in Illinois State, which is considered to be due to over-fitting as complexity of the model. That is, the classification performance was relatively degraded due to heterogeneous patterns and noise effect of input data, although the training accuracy of 3D-CNN model was high. This result simply that a proper classification algorithms should be selected considering spatial characteristics of study areas. Also, a large amount of training samples is necessary to guarantee higher classification performance in CNN, particularly in 3D-CNN.

• Science of Emotion and Sensibility
• /
• v.24 no.1
• /
• pp.91-104
• /
• 2021

This study explored users' emotional states identified from the wake-up words -"Hi, KIA!"- using a machine learning algorithm considering the user interface of passenger cars' voice. We targeted four emotional states, namely, excited, angry, desperate, and neutral, and created a total of 12 emotional scenarios in the context of car driving. Nine college students participated and recorded sentences as guided in the visualized scenario. The wake-up words were extracted from whole sentences, resulting in two data sets. We used the soundgen package and svmRadial method of caret package in open source-based R code to collect acoustic features of the recorded voices and performed machine learning-based analysis to determine the predictability of the modeled algorithm. We compared the accuracy of wake-up words (60.19%: 22%~81%) with that of whole sentences (41.51%) for all nine participants in relation to the four emotional categories. Accuracy and sensitivity performance of individual differences were noticeable, while the selected features were relatively constant. This study provides empirical evidence regarding the potential application of the wake-up words in the practice of emotion-driven user experience in communication between users and the artificial intelligence system.

• Journal of Digital Convergence
• /
• v.15 no.4
• /
• pp.191-198
• /
• 2017

It is needed to support intelligent customized health information service for user convenience in PHR based Personal Health Care Service Platform. In this paper, we specify an ontology-based health data model for Personal Health Care Service Platform. We also design a knowledge search algorithm that can be used to figure out similar health record by applying machine learning and data mining techniques. Axis-based mining algorithm, which we proposed, can be performed based on axis-attributes in order to improve relevance of knowledge exploration and to provide efficient search time by reducing the size of candidate item set. And K-Nearest Neighbor algorithm is used to perform to do grouping users byaccording to the similarity of the user profile. These algorithms improves the efficiency of customized information exploration according to the user 's disease and health condition. It can be useful to apply the proposed algorithm to a process of inference in the Personal Health Care Service Platform and makes it possible to recommend customized health information to the user. It is useful for people to manage smart health care in aging society.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.42 no.1
• /
• pp.127-134
• /
• 2022

Recently, policies and research to prevent increasing construction accidents have been actively conducted in the domestic construction industry. In previous studies, the prediction model developed to prevent construction accidents mainly used only structured data, so various characteristics of construction sites are not sufficiently considered. Therefore, in this study, we developed a machine learning-based construction accident prediction model that enables the characteristics of construction sites to be considered sufficiently by using both structured and text-type unstructured data. In this study, 6,826 cases of construction accident data were collected from the Construction Safety Management Integrated Information (CSI) for machine learning. The Decision forest algorithm and the BERT language model were used to train structured and unstructured data respectively. As a result of analysis using both types of data, it was confirmed that the prediction accuracy was 95.41 %, which is improved by about 20 % compared to the case of using only structured data. Conclusively, the performance of the predictive model was effectively improved by using the unstructured data together, and construction accidents can be expected to be reduced through more accurate prediction.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2007.05a
• /
• pp.605-608
• /
• 2007

천식(Asthma)과 같은 복합질환(Complex Disease)의 원인과 작용 모델을 찾기 위해서 여러가지 통계적인 방법들과 기계 학습(Machine Learning)의 방법 등이 사용되고 있다. 본 연구에서는 유전 알고리즘을 이용하여 천식 환자와 대조군들을 분류할 수 있는 단일염기 다형성(SNP, Single Nucleotide Polymorphism)의 조합에 대하여 조사한다.