• Journal of the Chungcheong Mathematical Society
• /
• v.32 no.1
• /
• pp.149-156
• /
• 2019

This paper present the global generalized cross validation as the appropriate choice of the regularization parameter in the preconditioned Gl-LSQR method in solving image deblurring problems. The regularization parameter, chosen from the global generalized cross validation, with preconditioned Gl-LSQR method can give better reconstructions of the true image than other parameters considered in this study.

• Smart Structures and Systems
• /
• v.10 no.4_5
• /
• pp.427-442
• /
• 2012

An effective approach for updating finite element model is presented which can provide reliable estimates for structural updating parameters from identified operational modal data. On the basis of the dynamic perturbation method, an exact relationship between the perturbation of structural parameters such as stiffness change and the modal properties of the tested structure is developed. An iterative solution procedure is then provided to solve for the structural updating parameters that characterise the modifications of structural parameters at element level, giving optimised solutions in the least squares sense without requiring an optimisation method. A regularization algorithm based on the Tikhonov solution incorporating the generalised cross-validation method is employed to reduce the influence of measurement errors in vibration modal data and then to produce stable and reasonable solutions for the structural updating parameters. The Canton Tower benchmark problem established by the Hong Kong Polytechnic University is employed to demonstrate the effectiveness and applicability of the proposed model updating technique. The results from the benchmark problem studies show that the proposed technique can successfully adjust the reduced finite element model of the structure using only limited number of frequencies identified from the recorded ambient vibration measurements.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.19 no.9
• /
• pp.1759-1771
• /
• 1994

We proposed the iterative image restoration method based on the method of steepest descent with a regularization constraint for restoring the noisy motion-blurred images. The conventional method proposed by Jan Biemond et al, had drawback to amplify the additive noise and make ringing effects in the restored images by determining the value of regularization parameter experimentally from the degraded image to be restored without considering local information of the restored one. The method we proposed had a merit to suppress the noise amplification and restoration error by using the regularization parameter which estimate the value of it adaptively from each pixels of the image being restored in order to reduce the noise amplification and ringing effects efficiently. Also we proposed the termination rule to stop the iteration automatically when restored results approach into or diverse from the original solution in satisfaction. Through the experiments, proposed method showed better result not only in a MSE of 196 and 453 but also in the suppression of the noise amplification in the flat region compared with those proposed by Jan Biemond et al. of which MSE of 216 and 467 respectively when we used 'Lean' and 'Jaguar' images as original images.

• The KIPS Transactions:PartB
• /
• v.16B no.1
• /
• pp.1-6
• /
• 2009

It is very difficult to restore the images degraded by motion blur and additive noise. In conventional methods, regularization usually applies to all the images without considering local characteristics of the images. As a result, ringing artifacts appear in the edge regions and noise amplification is in the flat regions, as well. To solve these problems, we propose an adaptive iterative regularization method, using the way of regularization operator considering edge directions. In addition, we suggest an adaptive regularization parameter and an relaxation parameter. In conclusion, We have verified that the new method shows the suppression of the noise amplification in the flat regions, also does less ringing artifacts in the edge regions. Furthermore, it offers better images and improves the quality of ISNR, comparing with those of conventional methods.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.66 no.12
• /
• pp.1772-1781
• /
• 2017

In this study, the design methodology for alleviating the overfitting problem of Polynomial Neural Networks(PNN) is realized with the aid of two kinds techniques such as L2 regularization and Sum of Squared Coefficients (SSC). The PNN is widely used as a kind of mathematical modeling methods such as the identification of linear system by input/output data and the regression analysis modeling method for prediction problem. PNN is an algorithm that obtains preferred network structure by generating consecutive layers as well as nodes by using a multivariate polynomial subexpression. It has much fewer nodes and more flexible adaptability than existing neural network algorithms. However, such algorithms lead to overfitting problems due to noise sensitivity as well as excessive trainning while generation of successive network layers. To alleviate such overfitting problem and also effectively design its ensuing deep network structure, two techniques are introduced. That is we use the two techniques of both SSC(Sum of Squared Coefficients) and $L_2$ regularization for consecutive generation of each layer's nodes as well as each layer in order to construct the deep PNN structure. The technique of $L_2$ regularization is used for the minimum coefficient estimation by adding penalty term to cost function. $L_2$ regularization is a kind of representative methods of reducing the influence of noise by flattening the solution space and also lessening coefficient size. The technique for the SSC is implemented for the minimization of Sum of Squared Coefficients of polynomial instead of using the square of errors. In the sequel, the overfitting problem of the deep PNN structure is stabilized by the proposed method. This study leads to the possibility of deep network structure design as well as big data processing and also the superiority of the network performance through experiments is shown.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.24 no.4
• /
• pp.382-391
• /
• 2021

Unsupervised domain adaptation often gives impressive solutions to handle domain shift of data. Most of current approaches assume that unlabeled target data to train is abundant. This assumption is not always true in practices. To tackle this issue, we propose a general solution to solve the domain gap minimization problem without any target data. Our method consists of two regularization steps. The first step is a pixel regularization by arbitrary style transfer. Recently, some methods bring style transfer algorithms to domain adaptation and domain generalization process. They use style transfer algorithms to remove texture bias in source domain data. We also use style transfer algorithms for removing texture bias, but our method depends on neither domain adaptation nor domain generalization paradigm. The second regularization step is a feature regularization by feature alignment. Adding a feature alignment loss term to the model loss, the model learns domain invariant representation more efficiently. We evaluate our regularization methods from several experiments both on small dataset and large dataset. From the experiments, we show that our model can learn domain invariant representation as much as unsupervised domain adaptation methods.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.30 no.4
• /
• pp.353-362
• /
• 2012

3D building modeling from airborne Lidar without model regularization may cause positional errors or topological inconsistency in building models. Regularization of 3D building models, on the other hand, restricts the types of models which can be reconstructed. To resolve these issues, this paper modelled 3D buildings from airborne Lidar by building model regularization which considers more various types of buildings. Building points are first segmented into roof planes by clustering in feature space and segmentation in object space. Then, 3D building models are reconstructed by consecutive adjustment of planes, lines, and points to satisfy parallelism, symmetry, and consistency between model components. The experimental results demonstrated that the method could make more various types of 3d building models with regularity. The effects of regularization on the positional accuracies of models were also analyzed quantitatively.

• Structural Engineering and Mechanics
• /
• v.75 no.4
• /
• pp.415-424
• /
• 2020

Recently, many structural damage detection (SDD) methods have been proposed to monitor the safety of structures. As an important modal parameter, mode shape has been widely used in SDD, and the difference of vectors was adopted based on sensitivity analysis and mode shapes in the existing studies. However, amplitudes of mode shapes in different measured points are relative values. Therefore, the difference of mode shapes will be influenced by their amplitudes, and the SDD results may be inaccurate. Focus on this deficiency, a multi-strategy SDD method is proposed based on the included angle of vectors and sparse regularization in this study. Firstly, inspired by modal assurance criterion (MAC), a relationship between mode shapes and changes in damage coefficients is established based on the included angle of vectors. Then, frequencies are introduced for multi-strategy SDD by a weighted coefficient. Meanwhile, sparse regularization is applied to improve the ill-posedness of the SDD problem. As a result, a novel convex optimization problem is proposed for effective SDD. To evaluate the effectiveness of the proposed method, numerical simulations in a planar truss and experimental studies in a six-story aluminum alloy frame in laboratory are conducted. The identified results indicate that the proposed method can effectively reduce the influence of noises, and it has good ability in locating structural damages and quantifying damage degrees.

• The Korean Journal of Applied Statistics
• /
• v.29 no.6
• /
• pp.1117-1128
• /
• 2016

In genetic association studies with high-dimensional genomic data, regularization procedures based on penalized likelihood are often applied to identify genes or genetic regions associated with diseases or traits. A network-based regularization procedure can utilize biological network information (such as genetic pathways and signaling pathways in genetic association studies) with an outstanding selection performance over other regularization procedures such as lasso and elastic-net. However, network-based regularization has a limitation because cannot be applied to high-dimension genomic data with a group structure. In this article, we propose to combine data dimension reduction techniques such as principal component analysis and a partial least square into network-based regularization for the analysis of high-dimensional genomic data with a group structure. The selection performance of the proposed method was evaluated by extensive simulation studies. The proposed method was also applied to real DNA methylation data generated from Illumina Innium HumanMethylation27K BeadChip, where methylation beta values of around 20,000 CpG sites over 12,770 genes were compared between 123 ovarian cancer patients and 152 healthy controls. This analysis was also able to indicate a few cancer-related genes.

• Journal of the Korean Data and Information Science Society
• /
• v.21 no.3
• /
• pp.461-470
• /
• 2010

Semi supervised classification which is a method using labeled and unlabeled data has considerable attention in recent years. Among various methods the graph based manifold regularization is proved to be an attractive method. Least squares support vector machine is gaining a lot of popularities in analyzing nonlinear data. We propose a semi supervised classification algorithm using the least squares support vector machines. The proposed algorithm is based on the manifold regularization. In this paper we show that the proposed method can use unlabeled data efficiently.