• 정보처리학회논문지B
• /
• 제14B권5호
• /
• pp.369-376
• /
• 2007

이 논문은 최적의 분류기 앙상블 선택을 위한 혼합 유전 알고리즘을 제안한다. 혼합 유전 알고리즘은 단순 유전알고리즘의 미세 조정력을 보완하기 위해 지역 탐색 연산을 추가한 것이다. 혼합 유전 알고리즘의 우수성을 입증하기 위해 단순 유전 알고리즘과 혼합 유전 알고리즘 각각을 비교 실험하였다. 또한 혼합 유전 알고리즘의 지역 탐색 연산으로 두 가지 방법(SSO: 순차 탐색 연산, CSO: 조합 탐색 연산)을 제안한다. 비교 실험 결과는 혼합 유전 알고리즘이 단순 유전 알고리즘에 비해 해를 탐색하는 능력이 우수하였다. 또한 분류기들의 상관관계를 고려한 CSO 방법이 SSO 방법보다 더 우수하였다.

• 설비공학논문집
• /
• 제20권7호
• /
• pp.508-516
• /
• 2008

Fault detection and diagnosis(FDD) system is beneficial in equipment management by providing the operator with tools which can help find out a failure of the system. An experimental study has been performed on fault detection and diagnosis method for a water chiller. Bayes classifier, which is one of classical pattern classifiers, is adopted in deciding whether fault occurred or not. Failure modes in this study include refrigerant leakage, decrease in mass flow rate of the chilled water and cooling water, and sensor error of the cooling water inlet temperature. It is possible to detect and diagnose faults in this study by adopting FDD algorithm using only four parameters(compressor outlet temperature, chilled water inlet temperature, cooling water outlet temperature and compressor power consumption). Refrigerant leakage failure is detected at 20% of refrigerant leakage. When mass flow rate of the chilled and cooling water decrease more than 8% or 12%, FDD algorithm can detect the faults. The deviation of temperature sensor over $0.6^{\circ}C$ can be detected as fault.

• 대한임베디드공학회논문지
• /
• 제10권5호
• /
• pp.325-333
• /
• 2015

Vision based night-time vehicle detection has been an emerging research field in various advanced driver assistance systems(ADAS) and automotive vehicle as well as automatic head-lamp control. In this paper, we propose night-time vehicle detection method based on multi-class support vector machine(SVM) that consists of thresholding, labeling, feature extraction, and multi-class SVM. Vehicle light candidate blobs are extracted by local mean based thresholding following by labeling process. Seven geometric and stochastic features are extracted from each candidate through the feature extraction step. Each candidate blob is classified into vehicle light or not by multi-class SVM. Four different multi-class SVM including one-against-all(OAA), one-against-one(OAO), top-down tree structured and bottom-up tree structured SVM classifiers are implemented and evaluated in terms of vehicle detection performances. Through the simulations tested on road video sequences, we prove that top-down tree structured and bottom-up tree structured SVM have relatively better performances than the others.

• 한국HCI학회:학술대회논문집
• /
• 한국HCI학회 2008년도 학술대회 3부
• /
• pp.172-177
• /
• 2008

생체 인식은 개인의 고유한 생체 정보를 획득하여 개인 식별에 이용하는 기술로, 그중 얼굴 인식은 사용자의 편의성과 비강제성이라는 장점이 있는 응용기술로 평가 받고 있다. 본 논문에서는 얼굴인식 기술동향을 살펴보고 얼굴 영역 추출, 특정 추출, 매칭을 포함한 시스템에 대해 논한다. 얼굴 영역 추출에는 얼굴 형판 정합 방법과 얼굴 요소의 검출에 의한 방법을, 특정 추출에서는 PCA 와 LDA 등의 방법을, 그리고 매칭을 통한 인증 단계에서는 최근접 분류기를 소개한다. 다양한 얼굴 인식 기법들이 제시됨에 따라 공인된 성능 평가 방법이 필요하게 되는데, 대용량 표준 얼굴 DE의 구축과 얼굴 인식 성능 평가 방법 개발의 필요성을 제시한다. 향후 얼굴인식 시스템에서는 조명, 자세, 표정의 변화를 어떻게 보정하여 인식 할 것인가 하는 것이 연구되어야 할 핵심 분야로서 3차원 얼굴 영상 복원 기술을 통한 해결방법을 살펴본다.

• 한국지능시스템학회:학술대회논문집
• /
• 한국퍼지및지능시스템학회 1998년도 추계학술대회 학술발표 논문집
• /
• pp.527-531
• /
• 1998

This study introduce to the fuzzy membership filter to cancel a high frequency noise of welding current. And ART2 which has the competitive learning network classifiers the signal patterns for the filtered welding signal. A welding current possesses a specific pattern according to the existence or the size of a welding gap. These specific patterns result in different classification in comparison with an occasion for no welding gap. The patterns In each case of 1mm, 2mm, 3mm, and no welding gap are identified by the artificial neural network. These procedure is an off-line execution. In on-line execution, the identification model of neural network for the classified pattern is located on ahead of the welding plant. And when the welding current patterns pass through the neural network in the direction of feedforward. it is possible to recognize the existence or the size of a welding gap.

• 한국지능시스템학회:학술대회논문집
• /
• 한국퍼지및지능시스템학회 1998년도 추계학술대회 학술발표 논문집
• /
• pp.361-367
• /
• 1998

The main two applications of the Genetic Algorithms(GA) are the optimization and the machine learning. Machine Learning has two objectives that make the complex system learn its environment and produce the proper output of a system. The machine learning using the Genetic Algorithms is called GA machine learning or genetic-based machine learning (GBML). The machine learning is different from the optimization problems in finding the rule set. In optimization problems, the population of GA should converge into the best individual because optimization problems, the population of GA should converge into the best individual because their objective is the production of the individual near the optimal solution. On the contrary, the machine learning systems need to find the set of cooperative rules. There are two methods in GBML, Michigan method and Pittsburgh method. The former is that each rule is expressed with a string, the latter is that the set of rules is coded into a string. Th classifier system of Holland is the representative model of the Michigan method. The classifier systems arrange the strength of classifiers of classifier list using the message list. In this method, the real time process and on-line learning is possible because a set of rule is adjusted on-line. A classifier system has three major components: Performance system, apportionment of credit system, rule discovery system. In this paper, we solve the food search problem with the learning and evolution of an artificial ant using the learning classifier system.

• Journal of Information Technology Applications and Management
• /
• 제12권2호
• /
• pp.15-27
• /
• 2005

Text categorization systems generally use single words (unigrams) as features. A deceptively simple algorithm for improving text categorization is investigated here, an idea previously shown not to work. It is to identify useful word pairs (bigrams) made up of adjacent unigrams. The bigrams it found, while small in numbers, can substantially raise the quality of feature sets. The algorithm was tested on two pre-classified datasets, Reuters-21578 for English and Korea-web for Korean. The results show that the algorithm was successful in extracting high quality bigrams and increased the quality of overall features. To find out the role of bigrams, we trained the Na$\"{i}$ve Bayes classifiers using both unigrams and bigrams as features. The results show that recall values were higher than those of unigrams alone. Break-even points and F1 values improved in most documents, especially when documents were classified along the large classes. In Reuters-21578 break-even points increased by 2.1%, with the highest at 18.8%, and F1 improved by 1.5%, with the highest at 3.2%. In Korea-web break-even points increased by 1.0%, with the highest at 4.5%, and F1 improved by 0.4%, with the highest at 4.2%. We can conclude that text classification using unigrams and bigrams together is more efficient than using only unigrams.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송공학회 2013년도 추계학술대회
• /
• pp.54-57
• /
• 2013

최근 다양한 방송 및 영상 분야에서 사람의 행동을 인식하여는 연구들이 많이 이루어지고 있다. 영상은 다양한 형태를 가질 수 있기 때문에 제약된 환경에서 유용한 템플릿 방법들보다 특징점에 기반한 연구들이 실제 사용자 환경에서 더욱 관심을 받고 있다. 특징점 기반의 연구들은 영상에서 움직임이 발생하는 지점들을 찾아내어 이를 3차원 패치들로 생성한다. 이를 이용하여 영상의 움직임을 히스토그램에 기반한 descriptor(서술자)로 표현하고 학습기반의 판별기(classifier)로 최종적으로 영상 내에 존재하는 행동들을 인식하였다. 그러나 단일 판별기를 이용한 다양한 영상 인식을 수용하기에는 힘들다. 최근에 이를 개선하기 위하여 다중 판별기를 활용한 연구들이 영상 판별 및 물체 검출 영역에서 사용되고 있다. 따라서 본 논문에서는 행동 인식을 위하여 support vector machine과 spare representation을 이용한 decision-level fusion 방법을 제안하고자 한다. 제안된 논문의 방법은 영상에서 특징점 기반의 descriptor를 추출하고 이를 각각의 판별기를 통하여 판별 결과들을 획득한다. 이 후 학습단계에서 획득된 가중치를 활용하여 각 결과들을 융합하여 최종 결과를 도출하였다. 본 논문에 실험에서 제안된 방법은 기존의 융합 방법보다 높은 행동 인식 성능을 보여 주었다.

• Industrial Engineering and Management Systems
• /
• 제10권4호
• /
• pp.272-278
• /
• 2011

Decision tree as a classification tool is being used successfully in many areas such as medical diagnosis, customer churn prediction, signal detection and so on. The main advantage of decision tree classifiers is their capability to break down a complex structure into a collection of simpler structures, thus providing a solution that is easy to interpret. Since decision tree is a top-down algorithm using a divide and conquer induction process, there is a risk of reaching a local optimal solution. This paper proposes a procedure of optimally determining thresholds of the chosen variables for a decision tree using an adaptive particle swarm optimization (APSO). The proposed algorithm consists of two phases. First, we construct a decision tree and choose the relevant variables. Second, we find the optimum thresholds simultaneously using an APSO for those selected variables. To validate the proposed algorithm, several artificial and real datasets are used. We compare our results with the original CART results and show that the proposed algorithm is promising for improving prediction accuracy.

• 한국정보과학회:학술대회논문집
• /
• 한국정보과학회 2005년도 한국컴퓨터종합학술대회 논문집 Vol.32 No.1 (B)
• /
• pp.226-228
• /
• 2005

Predicting the cellular location of an unknown protein gives valuable information for inferring the possible function of the protein. For more accurate Prediction system, we need a good feature extraction method that transforms the raw sequence data into the numerical feature vector, minimizing information loss. In this paper we propose new methods of extracting underlying features only from the sequence data by computing pairwise sequence alignment scores. In addition, we use composition based features to improve prediction accuracy. To construct an SVM ensemble from separately trained SVM classifiers, we propose specificity based weighted majority voting . The overall prediction accuracy evaluated by the 5-fold cross-validation reached $88.53\%$ for the eukaryotic animal data set. By comparing the prediction accuracy of various feature extraction methods, we could get the biological insight on the location of targeting information. Our numerical experiments confirm that our new feature extraction methods are very useful forpredicting subcellular localization of proteins.