• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2017년도 춘계학술발표대회
• /
• pp.936-937
• /
• 2017

최근 딥러닝을 통하여 영상의 카테고리 분류를 응용한 행동 인식이 활발히 연구되고 있다. 그러나 행동 인식을 위한 기존 연구 방법은 높은 수준의 하드웨어 사양을 요구하며 행동 인식에 대한 학습에 많은 시간이 소모되는 문제점을 지니고 있다. 또한, 행동 인식 테스트 결과를 얻기 위해 많은 시간이 소모되며 딥러닝 특성상 적은 수의 학습 데이터는 overfitting 문제를 일으킨다. 본 연구에서는 이러한 문제점을 해결하고자 행동인식을 위한 학습시간과 테스트 시간 감소를 위해 미리 학습된 VGG 모델을 사용해 얻어낸 RGB 데이터의 특징만을 학습에 사용하고 적은 수의 데이터로 행동 인식 테스트 결과를 높이기 위하여 RGB 데이터 증대를 통해 기존의 행동인식 연구보다 학습시간과 행동인식 테스트에 소모되는 시간을 줄인 방법을 행동 인식에 적용하였다. 이 방법을 UCF50 Dataset 에 적용하여 98.13%의 행동인식에 관한 정확성을 확인하였다.

• 산업공학
• /
• 제10권2호
• /
• pp.15-28
• /
• 1997

Knowledge acquisition has been a major bottleneck in building expert systems. To ease the problems arising in knowledge acquisition, analog concept learning systems(ACLS) has been used. In this paper, in order to avoid the overfitting problem and secure a good performance, we propose the revised ACLS, which pruning methods -cost complexity, reduced error, pessimistic pruning and production rule- are incorporated into and apply them to the credit evaluation for Korean companies. The performances of the revised ACLS are evaluated in light of the prediction accuracy. To check the effect of the training data sampling on the performance, experiments are conducted using the different proportion of the training data. Experimental results show that the revised ACLS of combining cost complexity pruning with reduced error pruning performs best among original ACLS and other methods.

• 한국멀티미디어학회논문지
• /
• 제16권11호
• /
• pp.1338-1347
• /
• 2013

Face recognition is a science of automatically identifying individuals based their unique facial features. In order to avoid overfitting and reduce the computational reduce the computational burden, a new face recognition algorithm using PCA-fisher linear discriminant (PCA-FLD) and fuzzy radial basis function neural network (RBFNN) is proposed in this paper. First, face features are extracted by the principal component analysis (PCA) method. Then, the extracted features are further processed by the Fisher's linear discriminant technique to acquire lower-dimensional discriminant patterns, the processed features will be considered as the input of the fuzzy RBFNN. As a widely applied algorithm in fuzzy RBF neural network, BP learning algorithm has the low rate of convergence, therefore, an improved learning algorithm based on Levenberg-Marquart (L-M) for fuzzy RBF neural network is introduced in this paper, which combined the Gradient Descent algorithm with the Gauss-Newton algorithm. Experimental results on the ORL face database demonstrate that the proposed algorithm has satisfactory performance and high recognition rate.

• Smart Structures and Systems
• /
• 제15권4호
• /
• pp.1103-1120
• /
• 2015

A self-sensing magnetorheological (MR) damper with embedded piezoelectric force sensor has recently been devised to facilitate real-time close-looped control of structural vibration in a simple and reliable manner. The development and characterization of the self-sensing MR damper are presented based on experimental work, which demonstrates its reliable force sensing and controllable damping capabilities. With the use of experimental data acquired under harmonic loading, a nonparametric dynamic model is formulated to portray the nonlinear behaviors of the self-sensing MR damper based on NARX modeling and neural network techniques. The Bayesian regularization is adopted in the network training procedure to eschew overfitting problem and enhance generalization. Verification results indicate that the developed NARX network model accurately describes the forward dynamics of the self-sensing MR damper and has superior prediction performance and generalization capability over a Bouc-Wen parametric model.

• Journal of the Korean Data and Information Science Society
• /
• 제18권1호
• /
• pp.51-61
• /
• 2007

Ridge regression (RR), principal component regression (PCR) and partial least squares regression (PLS) are among popular regression methods for collinear data. While RR adds a small quantity called ridge constant to the diagonal of X'X to stabilize the matrix inversion and regression coefficients, PCR and PLS use latent variables derived from original variables to circumvent the collinearity problem. One problem of PCR and PLS is that they are very sensitive to overfitting. A new regression method is presented by combining RR and PCR and PLS, respectively, in a unified manner. It is intended to provide better predictive ability and improved stability for regression models. A real-world data from NIR spectroscopy is used to investigate the performance of the newly developed regression method.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송∙미디어공학회 2017년도 추계학술대회
• /
• pp.139-141
• /
• 2017

화재는 다른 재난보다 확산 속도가 빠르기 때문에 신속하고 정확한 감지와 지속적인 감시가 요구된다. 최근, 신속하고 정확한 화재 감지를 위해, CCTV(Closed-Circuit TeleVision)으로 획득한 이미지를 기계학습(Machine Learning)을 이용해 화재 발생 여부를 감지하는 화재 감지 시스템이 주목받고 있다. 본 논문에서는 기계학습의 기술 중 정확도가 가장 높은 딥러닝(Deep Learning)기반의 CCTV 화재 감지 시스템을 제안한다. 본 논문의 시스템은 딥러닝 기술 적용뿐만이 아니라, CCTV 이미지 전처리 과정을 보완함으로써 딥러닝에서의 미지 데이터(unseen data)의 낮은 분류 정확도 문제인 과적합(overfitting)문제를 해결하였다. 본 논문의 시스템은 약 80,000 개의 CCTV 이미지 데이터를 학습하여, 90% 이상의 화재 이미지 분류 정확도의 성능을 보여주었다.

• Asia pacific journal of information systems
• /
• 제15권2호
• /
• pp.51-63
• /
• 2005

There has been substantial research into the bankruptcy prediction. Many researchers used the statistical method in the problem until the early 1980s. Since the late 1980s, Artificial Intelligence(AI) has been employed in bankruptcy prediction. And many studies have shown that artificial neural network(ANN) achieved better performance than traditional statistical methods. However, despite ANN's superior performance, it has some problems such as overfitting and poor explanatory power. To overcome these limitations, this paper suggests a relatively new machine learning technique, support vector machine(SVM), to bankruptcy prediction. SVM is simple enough to be analyzed mathematically, and leads to high performances in practical applications. The objective of this paper is to examine the feasibility of SVM in bankruptcy prediction by comparing it with ANN, logistic regression, and multivariate discriminant analysis. The experimental results show that SVM provides a promising alternative to bankruptcy prediction.

• Journal of Ship and Ocean Technology
• /
• 제11권4호
• /
• pp.21-28
• /
• 2007

Recently the importance of the utilization of engineering data is gradually increasing. Engineering data contains the experiences and know-how of experts. Data mining technique is useful to extract knowledge or information from the accumulated existing data. This paper deals with generating optimal polynomials using genetic programming (GP) as the module of Data Mining system. Low order Taylor series are used to approximate the polynomial easily as a nonlinear function to fit the accumulated data. The overfitting problem is unavoidable because in real applications, the size of learning samples is minimal. This problem can be handled with the extended data set and function node stabilization method. The Data Mining system for the ship design based on polynomial genetic programming is presented.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송공학회 2009년도 IWAIT
• /
• pp.238-241
• /
• 2009

In order to design a good quantizer for the underlying distribution using a training sequence (TS), the traditional approach is seeking for the empirical minimum based on the empirical risk minimization principle. As the size of TS increases, we may obtain a good quantizer for the true distribution. However, if we have a relatively small TS, searching the empirical minimum for the TS causes the overfitting problem, which even worsens the performance of the trained codebook. In this paper, the performance of codebooks trained by small TSs is studied, and it is shown that a piecewise uniform codebook can be better than an empirically minimized codebook is.

• Communications for Statistical Applications and Methods
• /
• 제13권1호
• /
• pp.113-123
• /
• 2006

Normal mixture model(NMM) frequently used to cluster genes on microarray gene expression data. In this paper some of component means of NMM are modelled by a linear regression model so that its design matrix presents the pattern between sample classes in microarray matrix. This modelling for the component means by given design matrices certainly has an advantage that we can lead the clusters that are previously designed. However, it suffers from 'overfitting' problem because in practice genes often are highly dimensional. This problem also arises when the NMM restricted by the linear model for component-means is fitted. To cope with this problem, in this paper, the use of the factor analyzer NMM restricted by linear model is proposed to cluster genes. Also several design matrices which are useful for clustering genes are provided.