• Communications for Statistical Applications and Methods
• /
• v.11 no.1
• /
• pp.49-58
• /
• 2004

In this talk we proposed an asymptotic test for dimensionality in the latent variable model for probabilistic principal component analysis with missing values at random. Proposed algorithm is a sequential likelihood ratio test for an appropriate Normal latent variable model for the principal component analysis. Modified EM-algorithm is used to find MLE for the model parameters. Results from simulations and real data sets give us promising evidences that the proposed method is useful in finding necessary number of components in the principal component analysis with missing values at random.

• Proceedings of the Korean Information Science Society Conference
• /
• 2001.04b
• /
• pp.529-531
• /
• 2001

We present an expectation-maximization algorithm for principal component analysis via orthogonalization. The algorithm finds actual principal components, whereas previously proposed EM algorithms can only find principal subspace. New algorithm is simple and more efficient thant probabilistic PCA specially in noiseless cases. Conventional PCA needs computation of inverse of the covariance matrices, which makes the algorithm prohibitively expensive when the dimensions of data space is large. This EM algorithm is very powerful for high dimensional data when only a few principal components are needed.

• Communications for Statistical Applications and Methods
• /
• v.10 no.3
• /
• pp.1057-1068
• /
• 2003

Results of using principal component analysis prior to cluster analysis are compared with results from applying agglomerative clustering algorithm alone. The retrieval ability of the agglomerative clustering algorithm is improved by using principal components prior to cluster analysis in some situations. On the other hand, the loss in retrieval ability for the agglomerative clustering algorithms decreases, as the number of informative variables increases, where the informative variables are the variables that have distinct information(or, necessary information) compared to other variables.

• Communications for Statistical Applications and Methods
• /
• v.10 no.3
• /
• pp.1025-1036
• /
• 2003

Variable selection algorithm for principal component analysis using penalty function is proposed. We use the fact that usual principal component problem can be expressed as a maximization problem with appropriate constraints and we will add penalty function to this maximization problem. Simulated annealing algorithm is used in searching for optimal solutions with penalty functions. Comparisons between several well-known penalty functions through simulation reveals that the HARD penalty function should be suggested as the best one in several aspects. Illustrations with real and simulated examples are provided.

• Journal of the Korean Data and Information Science Society
• /
• v.15 no.4
• /
• pp.817-823
• /
• 2004

This paper describes a three dimensional motion recognition algorithm and a system which adopts the algorithm for non-contact human-computer interaction. From sequence of stereos images, five feature regions are extracted with simple color segmentation algorithm and then those are used for three dimensional locus calculation precess. However, the result is not so stable, noisy, that we introduce principal component analysis method to get more robust motion recognition results. This method can overcome the weakness of conventional algorithms since it directly uses three dimensional information motion recognition.

• Journal of the Korean Mathematical Society
• /
• v.57 no.2
• /
• pp.401-414
• /
• 2020

It is computationally expensive to compute principal components from scratch at every update or downdate when new data arrive and existing data are truncated from the data matrix frequently. To overcome this limitations, incremental principal component analysis is considered. Specifically, we present a sliding window based efficient incremental principal component computation from a covariance matrix which comprises of two procedures; simultaneous update and downdate of principal components, followed by the rank-one matrix update. Additionally we track the accurate decomposition error and the adaptive numerical rank. Experiments show that the proposed algorithm enables a faster execution speed and no-meaningful decomposition error differences compared to typical incremental principal component analysis algorithms, thereby maintaining a good approximation for the principal components.

• Proceedings of the Korean Statistical Society Conference
• /
• 2003.05a
• /
• pp.151-156
• /
• 2003

Variable selection algorithm for principal component analysis using penalized likelihood method is proposed. We will adopt a probabilistic principal component idea to utilize likelihood function for the problem and use HARD penalty function to force coefficients of any irrelevant variables for each component to zero. Consistency and sparsity of coefficient estimates will be provided with results of small simulated and illustrative real examples.

• Bulletin of the Korean Chemical Society
• /
• v.29 no.4
• /
• pp.833-841
• /
• 2008

Principal component-genetic algorithm-multiparameter linear regression (PC-GA-MLR) and principal component-genetic algorithm-artificial neural network (PC-GA-ANN) models were applied for prediction of melting point for 323 drug-like compounds. A large number of theoretical descriptors were calculated for each compound. The first 234 principal components (PC’s) were found to explain more than 99.9% of variances in the original data matrix. From the pool of these PC’s, the genetic algorithm was employed for selection of the best set of extracted PC’s for PC-MLR and PC-ANN models. The models were generated using fifteen PC’s as variables. For evaluation of the predictive power of the models, melting points of 64 compounds in the prediction set were calculated. Root-mean square errors (RMSE) for PC-GA-MLR and PC-GA-ANN models are 48.18 and $12.77{^{\circ}C}$, respectively. Comparison of the results obtained by the models reveals superiority of the PC-GA-ANN relative to the PC-GA-MLR and the recently proposed models (RMSE = $40.7{^{\circ}C}$). The improvements are due to the fact that the melting point of the compounds demonstrates non-linear correlations with the principal components.

• Communications for Statistical Applications and Methods
• /
• v.29 no.6
• /
• pp.721-733
• /
• 2022

In this paper we compare parameter estimation by Grassmann manifold optimization and sequential candidate set algorithm in a structured principal fitted component (PFC) model. The structured PFC model extends the form of the covariance matrix of a random error to relieve the limits that occur due to too simple form of the matrix. However, unlike other PFC models, structured PFC model does not have a closed form for parameter estimation in dimension reduction which signals the need of numerical computation. The numerical computation can be done through Grassmann manifold optimization and sequential candidate set algorithm. We conducted numerical studies to compare the two methods by computing the results of sequential dimension testing and trace correlation values where we can compare the performance in determining dimension and estimating the basis. We could conclude that Grassmann manifold optimization outperforms sequential candidate set algorithm in dimension determination, while sequential candidate set algorithm is better in basis estimation when conducting dimension reduction. We also applied the methods in real data which derived the same result.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.17 no.2
• /
• pp.252-258
• /
• 2007

In this paper, a face recognition algorithm system using Principal Component Analysis (PCA) is proposed. The algorithm recognized a person by comparing characteristics (features) of the face to those of known individuals of Intelligent Control Laboratory (ICONL) face database. Simulations are carried out to investigate the algorithm recognition performance, which classified the face as a face or non-face and then classified it as known or unknown one. Particularly, a Principal Components of Linear Discriminant Analysis (PCA + LDA) face recognition algorithm is also proposed in order to confirm the recognition performances and the adaptability of a proposed PCA for a certain specific system.