• Journal of Korea Multimedia Society
• /
• v.24 no.1
• /
• pp.13-20
• /
• 2021

A Wasserstein GAN(WGAN), optimum in terms of minimizing Wasserstein distance, still suffers from inconsistent convergence or unexpected output due to inherent learning instability. It is widely known some kinds of restriction on the discriminative function should be considered to solve such problems, which implies the importance of Lipschitz continuity. Unfortunately, there are few known methods to satisfactorily maintain the Lipschitz continuity of the discriminative function. In this paper we propose techniques to stably maintain the Lipschitz continuity of the discriminative function by adding effective regularization terms to the objective function, which limit the magnitude of the gradient vectors of the discriminator to one or less. Extensive experiments are conducted to evaluate the performance of the proposed techniques, which shows the single-sided penalty improves convergence compared with the gradient penalty at the early learning process, while the proposed additional penalty increases inception scores by 0.18 after 100,000 number of learning.

• Journal of applied mathematics & informatics
• /
• v.35 no.3_4
• /
• pp.399-411
• /
• 2017

We first provide the linear operator equations corresponding to the Tikhonov regularization image denoising problems with different regularization terms, and then we propose how to choose Kronecker product preconditioners which are required for accelerating the $G{\ell}$-PCG method. Next, we provide how to apply the $G{\ell}$-PCG method with Kronecker product preconditioner to the linear operator equations. Lastly, we provide numerical experiments for image denoisng problems to evaluate the effectiveness of the $G{\ell}$-PCG with Kronecker product preconditioner.

• Journal of Information Processing Systems
• /
• v.13 no.5
• /
• pp.1168-1182
• /
• 2017

In the past decades, various image regularization methods have been introduced. Among them, total variation model has drawn much attention for the reason of its low computational complexity and well-understood mathematical behavior. However, regularization parameter estimation of total variation model is still an open problem. To deal with this problem, a novel adaptive regularization parameter selection scheme is proposed in this paper, by means of using the local spectral response, which has the capability of locally selecting the regularization parameters in a content-aware way and therefore adaptively adjusting the weights between the two terms of the total variation model. Experiment results on simulated and real noisy image show the good performance of our proposed method, in visual improvement and peak signal to noise ratio value.

• IE interfaces
• /
• v.24 no.2
• /
• pp.151-155
• /
• 2011

In this study, we propose a signomial classification method with ₀-regularization (₀-)which seeks a sparse signomial function by solving a mixed-integer program to minimize the weighted sum of the ₀-norm of the coefficient vector of the resulting function and the $L_1$-norm of loss caused by the function. $SC_0$ gives an explicit description of the resulting function with a small number of terms in the original input space, which can be used for prediction purposes as well as interpretation purposes. We present a practical implementation of $SC_0$ based on the mixed-integer programming and the column generation procedure previously proposed for the signomial classification method with $SL_1$-regularization. Computational study shows that $SC_0$ gives competitive performance compared to other widely used learning methods for classification.

• Communications for Statistical Applications and Methods
• /
• v.29 no.1
• /
• pp.41-51
• /
• 2022

We forecast the US oil consumption level taking advantage of google trends. The google trends are the search volumes of the specific search terms that people search on google. We focus on whether proper selection of google trend terms leads to an improvement in forecast performance for oil consumption. As the forecast models, we consider the least absolute shrinkage and selection operator (LASSO) regression and the structured regularization method for large vector autoregressive (VAR-L) model of Nicholson et al. (2017), which select automatically the google trend terms and the lags of the predictors. An out-of-sample forecast comparison reveals that reducing the high dimensional google trend data set to a low-dimensional data set by the LASSO and the VAR-L models produces better forecast performance for oil consumption compared to the frequently-used forecast models such as the autoregressive model, the autoregressive distributed lag model and the vector error correction model.

• Journal of the Korean Data and Information Science Society
• /
• v.22 no.1
• /
• pp.99-105
• /
• 2011

We propose a semi-supervised learning algorithm based on a form of regularization that incorporates similarity and dissimilarity penalty terms. Our approach uses a graph-based encoding of similarity and dissimilarity. We also present a model-selection method which employs cross-validation techniques to choose hyperparameters which affect the performance of the proposed method. Simulations using two types of dat sets demonstrate that the proposed method is promising.

• Journal of Information Processing Systems
• /
• v.15 no.6
• /
• pp.1449-1461
• /
• 2019

Face sketch synthesis plays an important role in public security and digital entertainment. In this paper, we present a novel face sketch synthesis method via local similarity and nonlocal similarity regularization terms. The local similarity can overcome the technological bottlenecks of the patch representation scheme in traditional learning-based methods. It improves the quality of synthesized sketches by penalizing the dissimilar training patches (thus have very small weights or are discarded). In addition, taking the redundancy of image patches into account, a global nonlocal similarity regularization is employed to restrain the generation of the noise and maintain primitive facial features during the synthesized process. More robust synthesized results can be obtained. Extensive experiments on the public databases validate the generality, effectiveness, and robustness of the proposed algorithm.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.24 no.1
• /
• pp.1-7
• /
• 2020

Fuzzy clustering, represented by FCM(Fuzzy C-Means), is a simple and efficient clustering method. However, the object function in FCM makes clusters affect clustering results proportional to the density of clusters, which can distort clustering results due to density difference between clusters. One method to alleviate this density problem is EDI-FCM(Extended Density-Independent FCM), which adds additional terms to the objective function of FCM to compensate for the density difference. In this paper, proposed is an enhanced EDI-FCM using regularization, Regularized EDI-FCM. Regularization is commonly used to make a solution space smooth and an algorithm noise insensitive. In clustering, regularization can reduce the effect of a high-density cluster on clustering results. The proposed method converges quickly and accurately to real centers when compared with FCM and EDI-FCM, which can be verified with experimental results.

• Steel and Composite Structures
• /
• v.39 no.5
• /
• pp.511-528
• /
• 2021

There are recently some advances in solving numerically topology optimization problems for large-scaled trusses based on ground structure approach. A disadvantage of this approach is that the final design usually includes many bars, which is difficult to be produced in practice. One of efficient tools is a so-called filter scheme for the ground structure to reduce this difficulty and determine several distinct bars. In detail, this technique is valuable for practical uses because unnecessary bars are filtered out from the ground structure to obtain a well-defined structure during the topology optimization process, while it still guarantees the global equilibrium condition. This process, however, leads to a singular system of equilibrium equations. In this case, the minimization of least squares with Tikhonov regularization is adopted. In this paper, a proposed algorithm in controlling optimal Tikhonov parameter is considered in combination with the filter scheme due to its crucial role in obtaining solution to remove numerical singularity and saving computational time by using sparse matrix, which means that the discrete optimal topology solutions depend on choosing the Tikhonov parameter efficiently. Several numerical examples are investigated to demonstrate the efficiency of the filter parameter control algorithm in terms of the large-scaled optimal topology designs.

• Journal of the Korean Mathematical Society
• /
• v.49 no.5
• /
• pp.927-945
• /
• 2012

An approximation of singular source terms in elliptic problems is developed and analyzed. Under certain assumptions on the curve where the singular source is defined, the second order convergence in the maximum norm can be proved. Numerical results present its better performance compared to previously developed regularization techniques.