• Journal of applied mathematics & informatics
• /
• v.29 no.5_6
• /
• pp.1245-1256
• /
• 2011

In this paper, we proposed a regularization interior point method for semidefinite programming with free variables which can be taken as an extension of the algorithm for standard semidefinite programming. Since an inexact search direction at each iteration is used, the computation of the designed algorithm is much less compared with the existing solution methods. The convergence analysis of the method is established under weak conditions.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.46 no.2
• /
• pp.78-88
• /
• 2009

In many face recognition problems, the number of available images is limited compared to the dimension of the input space which is usually equal to the number of pixels. This problem is called as the 'small sample size' problem and regularization methods are typically used to solve this problem in feature extraction methods such as LDA. By using regularization methods, the modified within class matrix becomes nonsingu1ar and LDA can be performed in its original form. However, in the process of adding a scaled version of the identity matrix to the original within scatter matrix, the scale factor should be set heuristically and the performance of the recognition system depends on highly the value of the scalar factor. By using the proposed resampling method, we can generate a set of images similar to but slightly different from the original image. With the increased number of images, the small sample size problem is alleviated and the classification performance increases. Unlike regularization method, the resampling method does not suffer from the heuristic setting of the parameter producing better performance.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2007.11a
• /
• pp.555-556
• /
• 2007

• Journal of Korea Multimedia Society
• /
• v.2 no.3
• /
• pp.320-328
• /
• 1999

The Laplacian operator is usually used as a regularization operator which may be used as any differential operator in the regularization iterative restoration. In this paper, several kinds of differential operator and 1-H operator that has been used in our lab as well, as a regularization operator, were compared with each other. In the restoration of noisy motion-blurred images, 1-H operator worked better than Laplacian operator in flat region, but in the edge the Laplacian operator operated better. For noisy gaussian-blurred image, 1-H operator worked better in the edge, while in flat region the Laplacian operator resulted better. In regularization, smoothing the noise and resorting the edges should be considered at the same time, so the regions divided into the flat, the middle, and the detailed, which were processed in separate and compared their MSE. Laplacian and 1-H operator showed to be suitable as the regularization operator, while the other differential operators appeared to be diverged as iterations proceeded.

• 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.

• Phonetics and Speech Sciences
• /
• v.7 no.3
• /
• pp.131-138
• /
• 2015

In this paper, we propose L1-norm regularization for state vector adaptation of subspace Gaussian mixture model (SGMM). When you design a speaker adaptation system with GMM-HMM acoustic model, MAP is the most typical technique to be considered. However, in MAP adaptation procedure, large number of parameters should be updated simultaneously. We can adopt sparse adaptation such as L1-norm regularization or sparse MAP to cope with that, but the performance of sparse adaptation is not good as MAP adaptation. However, SGMM does not suffer a lot from sparse adaptation as GMM-HMM because each Gaussian mean vector in SGMM is defined as a weighted sum of basis vectors, which is much robust to the fluctuation of parameters. Since there are only a few adaptation techniques appropriate for SGMM, our proposed method could be powerful especially when the number of adaptation data is limited. Experimental results show that error reduction rate of the proposed method is better than the result of MAP adaptation of SGMM, even with small adaptation data.

• Journal of the Institute of Convergence Signal Processing
• /
• v.5 no.3
• /
• pp.173-180
• /
• 2004

In this paper we provide a new image restoration method based on the multiscale regularization in the redundant wavelet transform domain. The proposed method uses the redundant wavelet transform to decompose the single-scale image restoration problem to multiscale ones and applies scale dependent regularization to the decomposed restoration problems. The proposed method recovers sharp edges by applying rather less regularization to wavelet related restorations, while suppressing the resulting noise magnification by the wavelet shrinkage algorithm. The improved performance of the proposed method over more traditional Wiener filtering is shown through numerical experiments.

• Journal of the Chungcheong Mathematical Society
• /
• v.27 no.4
• /
• pp.675-688
• /
• 2014

In this paper, we present an efficient way to determine a suitable value of the regularization parameter using the global generalized cross validation and analyze the experimental results from preconditioned global conjugate gradient linear least squares(Gl-CGLS) method in solving image deblurring problems. Preconditioned Gl-CGLS solves general linear systems with multiple right-hand sides. It has been shown in [10] that this method can be effectively applied to image deblurring problems. The regularization parameter, chosen from the global generalized cross validation, with preconditioned Gl-CGLS method can give better reconstructions of the true image than other parameters considered in this study.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.18 no.10
• /
• pp.75-80
• /
• 2017

In this paper, we describe the effect of the regularization method and the network with identity mapping on the performance of the activation functions in deep convolutional neural networks. The activation functions act as nonlinear transformation. In early convolutional neural networks, a sigmoid function was used. To overcome the problem of the existing activation functions such as gradient vanishing, various activation functions were developed such as ReLU, Leaky ReLU, parametric ReLU, and ELU. To solve the overfitting problem, regularization methods such as dropout and batch normalization were developed on the sidelines of the activation functions. Additionally, data augmentation is usually applied to deep learning to avoid overfitting. The activation functions mentioned above have different characteristics, but the new regularization method and the network with identity mapping were validated only using ReLU. Therefore, we have experimentally shown the effect of the regularization method and the network with identity mapping on the performance of the activation functions. Through this analysis, we have presented the tendency of the performance of activation functions according to regularization and identity mapping. These results will reduce the number of training trials to find the best activation function.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.9
• /
• pp.4658-4679
• /
• 2017

Android is now the most popular smartphone platform and remains rapid growth. There are huge number of sensitive privacy information stored in Android devices. Kinds of methods have been proposed to detect Android malicious applications and protect the privacy information. In this work, we focus on extracting the fine-grained features to maximize the information of Android malware detection, and selecting the least joint features to minimize the number of features. Firstly, permissions and APIs, not only from Android permissions and SDK APIs but also from the developer-defined permissions and third-party library APIs, are extracted as features from the decompiled source codes. Secondly, feature selection methods, including information gain (IG), regularization and particle swarm optimization (PSO) algorithms, are used to analyze and utilize the correlation between the features to eliminate the redundant data, reduce the feature dimension and mine the useful joint features. Furthermore, regularization and PSO are integrated to create a new joint feature mining method. Experiment results show that the joint feature mining method can utilize the advantages of regularization and PSO, and ensure good performance and efficiency for Android malware detection.