• Journal of Digital Convergence
• /
• v.19 no.10
• /
• pp.253-263
• /
• 2021

Machine learning is the process of constructing a cost function using learning data used for learning and an artificial neural network to predict the data, and finding parameters that minimize the cost function. Parameters are changed by using the gradient-based method of the cost function. The more complex the digital signal and the more complex the problem to be learned, the more complex and deeper the structure of the artificial neural network. Such a complex and deep neural network structure can cause over-fitting problems. In order to avoid over-fitting, a weight decay regularization method of parameters is used. We additionally use the value of the cost function in this method. In this way, the accuracy of machine learning is improved, and the superiority is confirmed through numerical experiments. These results derive accurate values for a wide range of artificial intelligence data through machine learning.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.25 no.1
• /
• pp.27-35
• /
• 2021

This paper proposes a method to mitigate the Catastrophic Forgetting phenomenon in which artificial neural networks forget information on previous data. This method adjusts the Regularization strength by measuring the relationship between previous data and present data. MNIST and EMNIST data were used for performance evaluation and experimented in three scenarios. The experiment results showed a 0.1~3% improvement in the accuracy of the previous task for the same domain data and a 10~13% improvement in the accuracy of the previous task for different domain data. When continuously learning data with various domains, the accuracy of all previous tasks achieved more than 50% and the average accuracy improved by about 7%. This result shows that neural network learning can be properly performed in a CL environment in which data of different domains are successively entered by the method of this paper.

• Proceedings of the Korean Fiber Society Conference
• /
• 2003.04a
• /
• pp.101-102
• /
• 2003

선형점탄성에서 완화시간분포를 알면 완화탄성율, 동적 점탄성율 등 다양한 정보를 알 수 있기 때문에 중요한 정보이지만 실험으로 직접 측정되는 물리량이 아니며 완화탄성율이나 동적 점탄성율의 실험 결과로부터 얻는 것도 많은 수학적인 어려움이 있다. 최근에 Regularization을 이용한 방법으로 연속함수로써 완화시간분포를 계산하는 방법들이 개발되어진 바 있다. 동적점탄성율 실험결과로부터 완화시간분포를 연속함수로써 계산할 수 있는 간단한 수치해법을 연구하였다. (중략)

• Bulletin of the Korean Mathematical Society
• /
• v.55 no.1
• /
• pp.319-330
• /
• 2018

This paper is concerned with the global asymptotic stability of strong solutions for a class of semilinear evolution equations with nonlocal initial conditions on infinite interval. The discussion is based on analytic semigroups theory and the gradually regularization method. The results obtained in this paper improve and extend some related conclusions on this topic.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.25 no.3
• /
• pp.252-261
• /
• 1988

Employed the new invese scattering scheme based on the moment mehtod, which was presented in the Part I of these companion papers, numerical simulations are performed to investigate the effect of measurement errors and noise contaminating the field scattered from dielectric objects. In order to reduce those effects on the reconstructed permittivity profiles, some techniques such as regularization, iterative matrix inversion, and multiple incidence are applied to this problem.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.20 no.9
• /
• pp.2887-2893
• /
• 1996

Governing equations infrictional contact problems are introduced using variational inequality formulations which are regularized to overcome the diffculties of non-differentiability of the friction functional. Also finite element approximations and a priori error estimations are derived based on those formulations. Numerical simulations are performed illustrating the theoretical results.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.22 no.5
• /
• pp.881-888
• /
• 1997

The existing method s for curvature estimation have a common problem in determining a unique smoothong factor. we previously proposed two approaches to overcome that problem: a constrained regularization approach and a mean field annealing approach. We consistently detected corners from the perprocessed smooth boundary obtained by either the constrained eglarization approach or the mean field annealing approach. Moreover, we defined corner sharpness to increase the robustness of both approaches. We evaluate the performance of those methods proposed in this paper. In addition, we show some matching results using a two-dimensional Hopfield neural network in the presence of occlusion as a demonstration of the power of our proposed methods.

• Communications of the Korean Mathematical Society
• /
• v.39 no.2
• /
• pp.513-534
• /
• 2024

The main goal of this work is to study an initial boundary value problem relating to the unsteady flow of a rigid, viscoplastic, and incompressible Bingham fluid in an elastic bounded domain of ℝ². By using the approximation sequences of the Faedo-Galerkin method together with the regularization techniques, we obtain the results of the existence and uniqueness of local solutions.

• Journal of applied mathematics & informatics
• /
• v.31 no.3_4
• /
• pp.353-363
• /
• 2013

A variant of the global conjugate gradient method for solving general linear systems with multiple right-hand sides is proposed. This method is called as the global conjugate gradient linear least squares (Gl-CGLS) method since it is based on the conjugate gradient least squares method(CGLS). We present how this method can be implemented for the image deblurring problems with Neumann boundary conditions. Numerical experiments are tested on some blurred images for the purpose of comparing the computational efficiencies of Gl-CGLS with CGLS and Gl-LSQR. The results show that Gl-CGLS method is numerically more efficient than others for the ill-posed problems.

• Knowledge Management Research
• /
• v.22 no.4
• /
• pp.27-40
• /
• 2021

Click-Through Rate (CTR) prediction is a key function that determines the ranking of candidate items in the recommendation system and recommends high-ranking items to reduce customer information overload and achieve profit maximization through sales promotion. The fields of natural language processing and image classification are achieving remarkable growth through the use of deep neural networks. Recently, a transformer model based on an attention mechanism, differentiated from the mainstream models in the fields of natural language processing and image classification, has been proposed to achieve state-of-the-art in this field. In this study, we present a method for improving the performance of a transformer model for CTR prediction. In order to analyze the effect of discrete and categorical CTR data characteristics different from natural language and image data on performance, experiments on embedding regularization and transformer normalization are performed. According to the experimental results, it was confirmed that the prediction performance of the transformer was significantly improved when the L2 generalization was applied in the embedding process for CTR data input processing and when batch normalization was applied instead of layer normalization, which is the default regularization method, to the transformer model.