• Korean Journal of Optics and Photonics
• /
• v.19 no.1
• /
• pp.9-14
• /
• 2008

In this paper, a new type of Yagi-Uda antenna that operates in the terahertz frequencies is designed. The proposed Yagi-Uda antenna can obtain high input resistance of approximately $2000{\Omega}$ at the resonance frequency by using a full-wavelength dipole instead of a half-wavelength dipole as the driver element. The current leakage into the bias line was minimized by applying the photonic bandgap structure to the bias line. By designing the antenna on a thin substrate, the impedance lowering of an antenna caused by the relative dielectric constant of the substrate was prevented and the end-fire radiation pattern which is the original radiation characteristic of the Yagi-Uda antenna could be obtained. We expect that the proposed Yagi-Uda antenna can achieve increased terahertz output power by improving the impedance mismatching problem with the photomixer.

• 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 IKEEE
• /
• v.27 no.3
• /
• pp.251-257
• /
• 2023

In this paper, we propose a dual-structured self-attention method that improves the lack of regional features of the vision transformer's self-attention. Vision Transformers, which are more computationally efficient than convolutional neural networks in object classification, object segmentation, and video image recognition, lack the ability to extract regional features relatively. To solve this problem, many studies are conducted based on Windows or Shift Windows, but these methods weaken the advantages of self-attention-based transformers by increasing computational complexity using multiple levels of encoders. This paper proposes a dual-structure self-attention using self-attention and neighborhood network to improve locality inductive bias compared to the existing method. The neighborhood network for extracting local context information provides a much simpler computational complexity than the window structure. CIFAR-10 and CIFAR-100 were used to compare the performance of the proposed dual-structure self-attention transformer and the existing transformer, and the experiment showed improvements of 0.63% and 1.57% in Top-1 accuracy, respectively.

• Economic Analysis
• /
• v.17 no.4
• /
• pp.30-51
• /
• 2011

Since the previous studies on the effects of trade liberalization implicitly assume that trade liberalization affects economic performance only in any point in time, they inevitably are static. Static evaluations fail to account for cumulative dynamic effects of trade liberalization that affect continuously economic performance. This paper tries to fill this gap of the previous studies in this field, estimating cumulative effects of trade liberalization on economic performance by employing an dynamic version of empirical model. One of important empirical issue is controlling bias from endogeneity. To resolve this problem, this paper employes system GMM that uses lagged first-differences as instruments for level equations and lagged levels as instruments for first-differences equations. It improves upon cross-section estimators because it controls for the potential bias induced by the omission of industry-specific effects and the endogeneity of all regressors. This study investigates the effects of trade liberalization in Korean manufacturing for the period from 1988 to 2005 and finds that cumulative dynamic effects of trade liberalization are present and bigger than static effects.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.19 no.12
• /
• pp.1375-1382
• /
• 2008

Frequency adjustable D-CRLH(dual-composite right/left handed) transmission lines, which solve the problem of design complexity and uncontrolled frequency of the existing structures, are proposed in this paper. The first design(type I), consisting of defected ground structure(DGS), island pattern in DGS, fixed stub and varactor diodes, controls $C_L$ in the parallel resonant circuit, while the second structure(type 2) composed of fixed DGS, shunt stub and diode adjusts $C_R$ in the series resonant circuit. The dual band frequency points which correspond to the meaningful electrical length of +/-90 degree in the RH/LH region are adjustable according to the bias voltage. The measurement shows that the LH frequency point which has -90 degree of electrical length are adjusted over $4.22{\sim}5.39\;GHz$ and $4.21{\sim}5.05\;GHz$ for type 1 and type 2, respectively, under $1{\sim}12\;V$ of bias voltage. In addition, the frequency Woo where RH turns over LH is controled over $3.26{\sim}4.22\;GHz$ for type 2 with the same bias condition.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.31 no.5B
• /
• pp.431-438
• /
• 2011

This study investigated the problem of mean-field bias correction under the assumption that the radar and rain gauge rainfall data follow the log-normal distribution. Regression curves for the average, median and mode of the radar and rain gauge rainfall were derived and evaluated for their usefulness. Additionally, these regression curves were compared with those derived under the assumption that the radar and rain gauge data follow the normal distribution. This study investigated the regression results for the Typhoon Meami occurred in 2003 as an example. As results, three regression lines with the radar rainfall as the independent variable were found to underestimate the rainfall, while those with the rain gauge rainfall as the independent variable to overestimate. Among three types of regression curves considered, the result for the average was most appropriate. However this case was found to be inferior to the regression line passing the origin under the assumption of the normal distribution with the rain gauge rainfall as its independent variable. So it was hard to conclude that the consideration of the log-normality on the correction of radar rainfall is beneficial.

• Korean Journal of Remote Sensing
• /
• v.32 no.5
• /
• pp.435-452
• /
• 2016

Land Surface Temperature (LST) has been operationally retrieved from the Communication, Ocean, and Meteorological Satellite (COMS) data by the spilt-window method (CSW_v2.0) developed by Cho et al. (2015). Although the CSW_v2.0 retrieved the LST with a reasonable quality compared to the Moderate Resolution Imaging Spectroradiometer (MODIS) LST data, it showed a relatively poor performance for the strong inversion and lapse rate conditions. To solve this problem, the LST retrieval algorithm (CSW_v2.0) was updated using the simulation results of radiative transfer model (MODTRAN 4.0) by considering the diurnal variations of air temperature. In general, the upgraded version, CSW_v3.0 showed a similar correlation coefficient between the prescribed LSTs and retrieved LSTs (0.99), the relatively smaller bias (from -0.03 K to-0.012 K) and the Root Mean Square Error (RMSE) (from 1.39 K to 1.138 K). Particularly, CSW_v3.0 improved the systematic problems of CSW_v2.0 that were encountered when temperature differences between LST and air temperature are very large and/or small (inversion layers and superadiabatic lapse rates), and when the brightness temperature differences and surface emissivity differences were large. The bias and RMSE of CSW_v2.0 were reduced by 10-30% in CSW_v3.0. The indirect validation results using the MODIS LST data showed that CSW_3.0 improved the retrieval accuracy of LST in terms of bias (from -0.629 K to -0.049 K) and RMSE (from 2.537 K to 2.502 K) compared to the CSW_v2.0.

• Journal of Educational Research in Mathematics
• /
• v.23 no.2
• /
• pp.75-94
• /
• 2013

As part of the education in the pursuit of equity, in this study, we have analyzed the differential item functioning on mathematics assessment through the result of 2011 National Assessment Educational Achievement. For this we used SIBTEST method and M-H method to extract differential item functioning on multicultural and North Korea migrant families students. As a result, 10 items that has the differential functioning were extracted by both methods in three school levels from Elementary, Middle and High School. The result of a exploratory for potential causes of differential functioning on multicultural and North Korea migrant families students through a qualitative analysis of each items that has been extracted, language ability, the complexity of computation and problem-solving process, the curriculum, the problem situation have been discussed. These results will be able to contribute to establishing education policy and designing teaching and learning methods for the multicultural and North Korea migrant families students.

• The Korean Journal of Financial Management
• /
• v.26 no.1
• /
• pp.189-220
• /
• 2009

This study tests the significance of momentum and contrarian strategy which challenge the weak efficient market hypothesis (EMH). If momentum and contrarian strategy can make extra return above the market, this can be a significant critics to the weak EMH. By using Monte Carlo simulation we have found that many existing returature, which test the significance of momentum and contrarian strategy, have a significance distortion problem. We test the significance of momentum and contrarian strategy by using reality check test of White(2000) which solve the problem of data snooping bias. The results are following. When we use the KOSPI index as the benchmark portfolio, we can get the best strategy of momentum strategy in the case of mean return. But in the case of Sharp ratio which is the performance measure adjusting risk, we find that the best strategy in the momentum and contrarian strategy can not dominate the performance of benchmark portfolio. Therefore we argue that weak EMH can not be rejected because of superior performance of momentum and contrarian strategy when we consider risk.

• Journal of Information Technology Services
• /
• v.18 no.2
• /
• pp.143-159
• /
• 2019

Machine learning (ML) is a method of fitting given data to a mathematical model to derive insights or to predict. In the age of big data, where the amount of available data increases exponentially due to the development of information technology and smart devices, ML shows high prediction performance due to pattern detection without bias. The feature engineering that generates the features that can explain the problem to be solved in the ML process has a great influence on the performance and its importance is continuously emphasized. Despite this importance, however, it is still considered a difficult task as it requires a thorough understanding of the domain characteristics as well as an understanding of source data and the iterative procedure. Therefore, we propose methods to apply deep learning for solving the complexity and difficulty of feature extraction and improving the performance of ML model. Unlike other techniques, the most common reason for the superior performance of deep learning techniques in complex unstructured data processing is that it is possible to extract features from the source data itself. In order to apply these advantages to the business problems, we propose deep learning based methods that can automatically extract features from transaction data or directly predict and classify target variables. In particular, we applied techniques that show high performance in existing text processing based on the structural similarity between transaction data and text data. And we also verified the suitability of each method according to the characteristics of transaction data. Through our study, it is possible not only to search for the possibility of automated feature extraction but also to obtain a benchmark model that shows a certain level of performance before performing the feature extraction task by a human. In addition, it is expected that it will be able to provide guidelines for choosing a suitable deep learning model based on the business problem and the data characteristics.