• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.11a
• /
• pp.244-247
• /
• 2020

Data augmentation has been helpful in improving the performance in deep learning, when we have a limited data and random erasing is one of the augmentations that have shown impressive performance in deep learning in multiple domains. But the main issue is that sometime it loses good features when randomly selected region is erased by some random values, that does not improve performance as it should. We target that problem in way that good features should not be lost and also want random erasing at the same time. For that purpose, we introduce new augmentation technique named Intra-Class Random Erasing (ICRE) that focuses on data to learn robust features of the same class samples by randomly exchanging randomly selected region. We perform multiple experiments by using different models including resnet18, VGG16 over variety of the datasets including ESC10, UrbanSound8K. Our approach has shown effectiveness over others methods including random erasing.

• International Journal of Reliability and Applications
• /
• v.1 no.1
• /
• pp.39-47
• /
• 2000

In a resent paper, Na and Kim(2000) develop a family of test statistics for testing whether or not the mean residual life changes its trend based on complete data and show that the new tests perform better than previously known tests. In this paper, we extend their tests to the randomly censored data. The asymptotic normality of the test statistics is established. Monte Carlo simulations are conducted to compare our tests with a previously known test by the power of tests.

• East Asian mathematical journal
• /
• v.34 no.3
• /
• pp.265-285
• /
• 2018

The main purpose of this paper, by using the resolvent operator technique associated with randomly (A, ${\eta}$, m)-accretive operator is to establish an existence and convergence theorem for a class of system of random nonlinear equations with fuzzy mappings in Banach spaces. Our works are improvements and generalizations of the corresponding well-known results.

• Journal of the Chungcheong Mathematical Society
• /
• v.15 no.1
• /
• pp.87-94
• /
• 2002

Trot(1999) considered how to calculate the expected area of a random triangle in the unit square $[0,1]{\times}[0,1]$. He used the Mathematica software package for the computational part. In this article, we study various aspects of the probability distribution of a triangle randomly chosen inside the circle of radius r. A simulation activity that can be conducted in statistics and probability classrooms is also considered.

• Journal of the Korean Data and Information Science Society
• /
• v.16 no.4
• /
• pp.1045-1052
• /
• 2005

This paper deals with the estimations of kernel ridge regression when the responses are subject to randomly right censoring. The weighted data are formed by redistributing the weights of the censored data to the uncensored data. Then kernel ridge regression can be taken up with the weighted data. The hyperparameters of model which affect the performance of the proposed procedure are selected by a generalized approximate cross validation(GACV) function. Experimental results are then presented which indicate the performance of the proposed procedure.

• Journal of the Korean Statistical Society
• /
• v.33 no.1
• /
• pp.25-34
• /
• 2004

In this paper we propose a prediction method on the regression model with randomly censored observations of the training data set. The least squares support vector machine regression is applied for the regression function prediction by incorporating the weights assessed upon each observation in the optimization problem. Numerical examples are given to show the performance of the proposed prediction method.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.61 no.2
• /
• pp.216-222
• /
• 2012

Newly proposed RG-MADS (Randomly Guided Mesh Adaptive Direct Search) has been applied to the optimal design of Interior Permanent Magnet Synchronous Motor (IPMSM) which has the distinctive features of magnetic saturation. RG-MADS, advanced from classical MADS algorithm, has the superiority in computational time and reliable convergence accuracy to the optimal solution, thus it is appropriate to the optimal design of IPMSM coupled with time-consuming Finite Element Analysis (FEA), necessary to the nonlinear magnetic application for better accuracy. Effectiveness of RG-MADS has been verified through the well-known benchmark-functions beforehand. In addition, the proposed RG-MADS has been applied to the optimal design of IPMSM aiming at maximizing the Maximum Torque Per Ampere (MTPA), which is regarded as representative design goal of IPMSM.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.62 no.8
• /
• pp.1132-1137
• /
• 2013

This paper considers the $H_{\infty}$ control problem for networked control systems(NCSs). In order to solve the problem which comes from discontinuous control signal in NCSs, an approach that discontinuous control signals treat time-varying delayed continuous signals is applied to achieve $H_{\infty}$ stability of NCSs. In addition, randomly occurring packet losses and disturbances are considered by introducing stochastic variables with Bernoulli distribution. Based on Lyapunov stability theory, a new stability condition is obtained via linear matrix inequality formulation to find the $H_{\infty}$ controller which achieves the mean square stability of NCSs. Finally, the proposed method is applied to a numerical example in order to show the effectiveness of our results.

• Journal of the Korean Institute of Electrical and Electronic Material Engineers
• /
• v.18 no.4
• /
• pp.338-343
• /
• 2005

Pb(Zr,Ti)O₃[PZT] thin films were prepared on a highly (100) oriented LaNiO₃[LNO] and a randomly oriented LNO by sol-gel process. The PZT thin films on a highly (100) oriented LNO show a high (100) crystal orientation (F=100 ％), those on a randomly oriented LNO show a random crystal orientation (F=60 ％). All the PZT layer have a flat and dense microstructure with large columnar grains and their grain size are 25 nm. In the ferroelectric curves at electric field of 40 kV/cm, a highly (100) oriented PZT/LNO samples show coercive field, E/sub c/=10 kV/cm and remanent polarization, P/sub r/=14.5 μC/㎠, while a randomly oriented PZT/LNO sample show E/sub c/=10 kV/cm and P/sub r/=5.4 μC/㎠.

• Journal of the Korean Data and Information Science Society
• /
• v.26 no.1
• /
• pp.77-87
• /
• 2015

Parrondo paradox is the counter-intuitive phenomenon where two losing games can be combined to win or two winning games can be combined to lose. In this paper, we consider an ensemble of players, one of whom is chosen randomly to play game A' or game B. In game A', the randomly chosen player transfers one unit of his capital to another randomly selected player. In game B, the player plays the history-dependent Parrondo game in which the winning probability of the present trial depends on the results of the last two trials in the past. We show that Parrondo paradox exists in this redistribution model of the history-dependent Parrondo game.