• The Journal of Society for e-Business Studies
• /
• v.25 no.4
• /
• pp.77-109
• /
• 2020

In this paper, we build a gradient Boosting model to predict government SME R&D subsidy, select features of high importance, and measure the impact of each features to the predicted subsidy using PDP and SHAP value. Unlike previous empirical researches, we focus on the effect of the R&D subsidy distribution pattern to the incentive of the firms participating subsidy competition. We used the firm data constructed by KISTEP linking government R&D subsidy record with financial statements provided by NICE, and applied a Gradient Boosting model to predict R&D subsidy. We found that firms with higher R&D performance and larger R&D investment tend to have higher R&D subsidies, but firms with higher operation profit or total asset turnover rate tend to have lower R&D subsidies. Our results suggest that current government R&D subsidy distribution pattern provides incentive to improve R&D project performance, but not business performance.

• Journal of Korean Society of Forest Science
• /
• v.86 no.1
• /
• pp.17-24
• /
• 1997

As the first step to estimate the potential natural vegetation distribution of the globe, the best spherical interpolation method was developed to the temperature and precipitation which have close relation to the distribution pattern of world natural vegetation. For developing the interpolation method, a named Light Climatic Dataset composed of 1,060 stations around the globe was randomly divided into halves of feeding side and target side. The discrepancy between the observed and estimated values at the target stations was compared with combinations of parameters and methods. The estimated values were calculated to each combination which is all-out, constant radius and constant station methods in the selection of the feeding stations, n square reciprocal and negative exponential functions in weighting function of distance between feeding stations and each target, and oval weighting in direction of the feeding stations from each target. As a result, it turned out that the spherical interpolation with negative exponential weighting function fed from the constant radius stations ovally weighed yields the best estimates both for temperature and for precipitation. The parameters for temperature are $30^{\circ}$ in constant radius, 0.78 in negative exponential function and 0.4 in oval weighting, and for precipitation are $30^{\circ}$, 0.53 and 0.4, respectively.

• Smart Media Journal
• /
• v.10 no.1
• /
• pp.9-15
• /
• 2021

The eye tracking system designed in this paper is an eye-based computer input device designed to give an easy access for those who are uncomfortable with Lou Gehrig's or various muscle-related diseases. It is an eye-based-computer-using device for users whose potential demand alone amounts to 30,000. Combining the number of Lou Gehrig's patients in Korea estimated at around 1,700, and those who are unable to move their bodies due to various accidents or diseases. Because these eye input devices are intended for a small group of users, many types of commercial devices are available on the market. It is making them more expensive and difficult to use for these potential users, less accessible. For this reason, each individual's economic situation and individual experience with smart devices are slightly different. Therefore, making it difficult to access them in terms of cost or usability to use a commercial eye tracking system. Accordingly, attempts to improve accessibility to IT devices through low-cost but easy-to-use technologies are essential. Thus, this paper proposes a complementary superior performance eye tracking system that can be conveniently used by far more people and patients by improving the deficiencies of the existing system. Through voluntary VoCs(Voice of Customers) of users who have used different kinds of eye tracking systems that satisfies it through various usability tests, and we propose a reduced system that the amount of calculation to 1/15th, and eye-gaze tracking error rate to 0.5~1 degree under.

• Journal of the Korean Geotechnical Society
• /
• v.24 no.9
• /
• pp.13-21
• /
• 2008

Partitioning tracer method is a useful tool to characterize large domains of the aquifers contaminated with light nonaqueous phase liquids (LNAPLs). Sorption of the partitioning tracers to the organic matter content of soil can potentially influence the efficacy of partitioning tracer method. LNAPL-water partitioning coefficients of tracers ($K_{nw}$), measured by static method, showed linear relationship. Sorption isotherm tests were conducted to evaluate the sorption capacity of the soils packed in the columns and the results were appropriately represented by Freundlich sorption isotherm. The sorption of tracers proportionally increased with the increase of the organic matter content of the soil. Laboratory experiments were conducted in four columns each packed with soils of different organic matter contents to determine the potential interference effects of sorption to soil organic matter content and correction factors for the errors in estimation of LNAPLs by partitioning tracer method. Though there were no contaminants added, breakthrough curves from columns packed with mixture of Jumunjin standard sand and organic matter showed separation of tracers. Columns were then contaminated to residual saturation with kerosene and breakthrough curves were obtained. The results show that sorption of tracers to soil organic matter leads to an increase in the retardation factor (R) and hence, to an overestimation of the saturation of LNAPLs. A relation between the percentage of organic matter content and the corresponding percentage error in the estimation of NAPLs has been developed.

• The Journal of the Acoustical Society of Korea
• /
• v.14 no.1E
• /
• pp.56-64
• /
• 1995

The least mean fourth (LMF) adaptive algorithm is a stochastic gradient method that minimizes the error in the mean fourth sense. Despite its potential advantages, the algorithm is much less popular than the conventional least mean square (LMS) algorithm in practice. This seems partly because the analysis of the LMF algorithm is much more difficult than that of the LMS algorithm, and thus not much still has been known about the algorithm. In this paper, we explore the statistical convergence behavior of the LMF algorithm when the input to the adaptive filter is zero-mean, wide-sense stationary, and Gaussian. Under a system idenrification mode, a set of nonlinear evolution equations that characterizes the mean and mean-squared behavior of the algorithm is derived. A condition for the conbergence is then found, and it turns out that the conbergence of the LMF algorithm strongly depends on the choice of initial conditions. Performances of the LMF algorithm are compared with those of the LMS algorithm. It is observed that the mean convergence of the LMF algorithm is much faster than that of the LMS algorithm when the two algorithms are designed to achieve the same steady-state mean-squared estimation error.

• The Korean Journal of Nuclear Medicine Technology
• /
• v.24 no.1
• /
• pp.15-19
• /
• 2020

Purpose Generative Adversarial Network(GAN) is one of deep learning technologies. This is a way to create a real fake image after learning the real image. In this study, after acquiring artificial intelligence images through GAN, We were compared and evaluated with real scan time images. We want to see if these technologies are potentially useful. Materials and Methods 30 patients who underwent ¹⁸F-FDG Brain PET/CT scanning at Severance Hospital, were acquired in 15-minute List mode and reconstructed into 1,2,3,4,5 and 15minute images, respectively. 25 out of 30 patients were used as learning images for learning of GAN and 5 patients used as verification images for confirming the learning model. The program was implemented using the Python and Tensorflow frameworks. After learning using the Pix2Pix model of GAN technology, this learning model generated artificial intelligence images. The artificial intelligence image generated in this way were evaluated as Mean Square Error(MSE), Peak Signal to Noise Ratio(PSNR), and Structural Similarity Index(SSIM) with real scan time image. Results The trained model was evaluated with the verification image. As a result, The 15-minute image created by the 5-minute image rather than 1-minute after the start of the scan showed a smaller MSE, and the PSNR and SSIM increased. Conclusion Through this study, it was confirmed that AI imaging technology is applicable. In the future, if these artificial intelligence imaging technologies are applied to nuclear medicine imaging, it will be possible to acquire images even with a short scan time, which can be expected to reduce artifacts caused by patient movement and increase the efficiency of the scanning room.

• Korean Journal of Remote Sensing
• /
• v.34 no.6_3
• /
• pp.1383-1398
• /
• 2018

Sea surface wind is one of the most fundamental variables for understanding diverse marine phenomena. Although scatterometers have produced global wind field data since the early 1990's, the data has been used limitedly in oceanic applications due to it slow spatial resolution, especially at coastal regions. Synthetic Aperture Radar (SAR) is capable to produce high resolution wind field data. KOMPSAT-5 is the first Korean satellite equipped with X-band SAR instrument and is able to retrieve the sea surface wind. This study presents the validation results of sea surface wind derived from the KOMPSAT-5 backscattering coefficient data for the first time. We collected 18 KOMPSAT-5 ES mode data to produce a matchup database collocated with buoy stations. In order to calculate the accurate wind speed, we preprocessed the SAR data, including land masking, speckle noise reduction, and ship detection, and converted the in-situ wind to 10-m neutral wind as reference wind data using Liu-Katsaros-Businger (LKB) model. The sea surface winds based on XMOD2 show root-mean-square errors of about $2.41-2.74m\;s^{-1}$ depending on backscattering coefficient conversion equations. In-depth analyses on the wind speed errors derived from KOMPSAT-5 backscattering coefficient data reveal the existence of diverse potential error factors such as image quality related to range ambiguity, discrete and discontinuous distribution of incidence angle, change in marine atmospheric environment, impacts on atmospheric gravity waves, ocean wave spectrum, and internal wave.

• Journal of Korean Society of Transportation
• /
• v.23 no.4 s.82
• /
• pp.47-55
• /
• 2005

In recent years, the rapid growth of vehicles have increased traffic crashes. Since they can cause the economic losses and have put the life qualify in danger, there should be numerous efforts to reduce traffic crashes. To reduce traffic crashes, this research seeks to improve the safety of intersections by analysing causations of injury severity with Ordered Probability Model. This research applied the Ordered Probit Model, which assumes that ${\epsilon}_i$(random error) is normally distributed, for model calibration and used $p^2$ (likelihood ratio) and $x^2$ (Chi-square) for model selection. The results show that minor road traffic, heavy vehicle rates, major and minor right-turn rates, presence of lightings, speed limits, instructive line for left-turn traffic are significant factors affecting crash severities at signalized intersections.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.19 no.12
• /
• pp.2844-2851
• /
• 2015

How to recover from the multiple 2D images into 3D object has been widely studied in the field of computer vision. In order to improve the accuracy of the recovered 3D shape, it is more important that noise must be minimized and the number of image frames must be guaranteed. However, potential noise is implied when tracking feature points. And the number of image frames which is consisted of an observation matrix usually decrease because of tracking failure, occlusions, or low image resolution, and so on. Therefore, it is obviously essential that the number of image frames must be secured by recovering the missing feature points under noise. Thus, we propose the analytic approach which can control directly the error distance and orientation of missing feature point by the geometrical properties under noise distribution. The superiority of proposed method is demonstrated through experimental results for synthetic and real object.

• Journal of Korea Soil Environment Society
• /
• v.3 no.3
• /
• pp.11-20
• /
• 1998

The effects of soil venting on the leaching potential of residual gasoline were characterized by applying a simple multi-component volatilization/dissolution model based on Raoult's law. The validity of Raoult's law in describing dissolution of gasoline was evaluated separately using both pure gasoline and gasoline contaminated soil. The aqueous concentrations of gasoline components equilibrated with pure gasoline could be described by Raoult's law within one order of magnitude, regardless of the composition of the gasoline. The leaching concentrations from contaminated soil could be well predicted at a relatively high gasoline concentration in soil. However, after 93.5% removal of gasoline by venting, the calculated values were higher than the experimental values by 50∼100%. A model involving multi-component evaporation and dissolution was applied and the results were compared with the experimental values. Possible causes of the discrepancy between the predicted values and experimental values were given.