• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.18 no.10
• /
• pp.75-80
• /
• 2017

In this paper, we describe the effect of the regularization method and the network with identity mapping on the performance of the activation functions in deep convolutional neural networks. The activation functions act as nonlinear transformation. In early convolutional neural networks, a sigmoid function was used. To overcome the problem of the existing activation functions such as gradient vanishing, various activation functions were developed such as ReLU, Leaky ReLU, parametric ReLU, and ELU. To solve the overfitting problem, regularization methods such as dropout and batch normalization were developed on the sidelines of the activation functions. Additionally, data augmentation is usually applied to deep learning to avoid overfitting. The activation functions mentioned above have different characteristics, but the new regularization method and the network with identity mapping were validated only using ReLU. Therefore, we have experimentally shown the effect of the regularization method and the network with identity mapping on the performance of the activation functions. Through this analysis, we have presented the tendency of the performance of activation functions according to regularization and identity mapping. These results will reduce the number of training trials to find the best activation function.

• Journal of the military operations research society of Korea
• /
• v.29 no.1
• /
• pp.28-42
• /
• 2003

The simplex method requires basis update in each iteration, which is the most time consuming process. Several methods have been developed for the update of basis which is represented in LU factorized form, such as Bartels-Golub's method, Forrest-Tomlin's method, Reid's method, Saunders's method, etc. In this research, we compare between the updating methods in terms of sparsity, data structure and computing time issues. The analysis is mainly based on the computational experience.

• The Journal of Engineering Geology
• /
• v.34 no.2
• /
• pp.279-294
• /
• 2024

Large-scale civil engineering structures such as dams require a systematic approach to jointed rock-mass grouting to prevent water leakage into the foundations and to ensure safe operation. In South Korea, rock grouting design often relies on the experience of field engineers that was gained in similar projects, highlighting the need for a more systematic and reliable approach. Rock-mass grouting is affected mainly by hydrogeology and the presence of discontinuities, involving factors such as the rock quality designation (RQD), joint spacing (Js), Lugeon value (Lu), and secondary permeability index (SPI). This study, based on data from field investigations of 14 domestic sites, analyzed the correlation between hydrogeological factors (Lu and SPI), discontinuity characteristics (RQD and Js), and grout take, and systematically established a design method for rock grouting. Analysis of correlation between the variables RQD, Js, Lu, and SPI yielded Pearson correlation (r) values as follows: Lu-SPI, 0.92; RQD-Lu, -0.75; RQD-Js, 0.69; RQD-SPI, -0.65; Js-Lu, -0.47; and SPI-Js, -0.41. The grout take increases with Lu and SPI values, but there is no significant correlation between RQD and Js. The proposed approach for grouting design based on SPI values was verified through analysis and comparison with actual curtain-grouting construction, and is expected to be useful in practical applications and future studies.

• Journal of Life Science
• /
• v.17 no.8 s.88
• /
• pp.1082-1089
• /
• 2007

Serological anlysis of recombinant cDNA expression libraries (SEREX) has led to identification of several categories of new antigens recognized by the immune system of cancer patients, which are referred to as the cancer immunome. We analyzed normal testis cDNA expression libraries with serumobtained from non-small lung cancer patient and isolated 40 distinct antigen designated KP-LuT-1 through KP-LuT-40. Among these antigens 20 antigens were previously identified by SEREX analysis of other tumor types, and 20 out of 40 antigens (50%) did not match entries in Cancer Immunome Database and were considered newly identified antigens. Sequencing analysis showed that the anti-gens comprised 26 functional known proteins and 14 noble/uncharacterized gene products. Of these, the hypothetical protein KP-LuT-6 was shown tissue-restricted. RT-PCR showed it to be expressed strongly only in normal testis. In addition to normal tissues-restricted expression, KP-LuT-6 mRNA was detected in lung tumor samples(3/l0), stomach tumor samples(3/l0), and breast tumor samples(l/5), whereas not detected in colon tumor samples(O/I2). These data suggest that KP-LuT-6 is a cancer/testis (CT)-like antigen as a potential target for cancer immunotherapies.

• Journal of the Korean Chemical Society
• /
• v.44 no.6
• /
• pp.541-546
• /
• 2000

Methods to determine L-dopa(L-3,4-dihydroxyphenylalanine) in aqueous solution by spectrofluorimetry based upon the ligand sensitized luminescence of Tb(III) ion L-dopa complex have been studied. Tb(III) ion and Lu(III) ion were used as ligand sensitized fluorescencer and co-fluorescence enhancer, respectively. The effects of excitation wavelength, pH, concentration of Tb(III) ion, concentration of Lu(III) ion and emission wavelength on the fluorescence intensity were investigated. The fluorescence intensity of the Tb(III) ion L-dopa complex was further increased with addition of Lu(III) ion. The calibration curve for L-dopa was linear over the range from 5.0 ${\times}$ $10^{-7}$ M to 1.0${\times}$ $10^{-4}$ M and the detection limit was 4.0 ${\times}$ $10^{-8}$ M under the optimal experimental conditions of 300 nm, 8.0, 1.0 ${\times}$ $10^{-4}$ M and 545 nm for excitation wavelength, pH, concentration of Tb(III) ion and emission wavelength, respectively. When Lu(III) ion was added to Tb(Ⅲ) ion L-dopa complex, the concentration range of linear response and detection limit were 1.0${\times}$$10^{-8}$ M to 2.0 ${\times}$ $10^{-4}$ M and 1.0 ${\times}$ $10^{-9}$ M, respectively under the optimal experimental conditions of 300 nm, 8.5, 1.0 ${\times}$ $10^{-5}$ M, 1.0 ${\times}$ $10^{-5}$ M, 545 nm for excitation wavelength, pH, concentration of Tb(III) ion, concentration of Lu(III) ion and emission wavelength, respectively.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.30 no.7
• /
• pp.68-80
• /
• 2002

An approach for composing a performance optimized computational code is suggested for latest microprocessors. The concept of the code optimization, called here as localization, is maximizing the utilization of the second level cache that is common to all the latest computer system, and minimizing the access to system main memory. In this study, the localized optimization of LU-SGS (Lower-Upper Symmetric Gauss-Seidel) code for the solution of fluid dynamic equations was carried out in three different levels and tested for several different microprocessor architectures most widely used in these days. The test results of localized optimization showed a remarkable performance gain up to 7.35 times faster solution, depending on the system, than the baseline algorithm for producing exactly the same solution on the same computer system.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.16 no.2
• /
• pp.287-292
• /
• 2021

In this paper, we propose an efficient CNN application method by analyzing the effect of activation function on the failure diagnosis of the inductive motor stator. Generally, the main purpose of the inductive motor stator failure diagnosis is to prevent the failure by rapidly diagnosing the minute turn short. In the application of activation function, experiments show that the Sigmoid function is 23.23% more useful in accuracy of diagnosis than the ReLu function, although it is shown that ReLu has superiority in overall fixer failure in utilizing the activation function.

• Journal of the Korean Society of Propulsion Engineers
• /
• v.8 no.1
• /
• pp.16-26
• /
• 2004

The objective of this study is to apply preconditioning to Wavier-Stokes equations with a turbulence model. The concept of a pseudo sonic speed was adopted. Roe's FDS was used for spatial discretization, LU-SGS scheme was used for time integration. In order to test the algorithms, the low speed flows around NACA airfoils and the flows through supersonic nozzle were calculated. The algorithm developed in the present study shows good performance in the calculations of low speed viscous flows and supersonics flows.

• The Journal of Internal Korean Medicine
• /
• v.36 no.3
• /
• pp.228-235
• /
• 2015

Objectives: The aim of this study was to find an effective medical device to detect cold sensation of hands. Methods: Patients with cold sensation of hands (N=42) were recruited in July 2013. We surveyed patients’ general characteristics, severity of cold sensation, body temperature in Nogung (PC8) and Hyeopbaek (LU4) measured with a digital thermometer, digital infrared thermal imaging, and pulse wave velocity. Results: The relation between severity of cold sensation and temperature in Nogung (PC8) and Hyeopbaek (LU4) was statistically significant. The results showed a higher correlation of severity of cold sensation with Nogung (PC9) than with Hyeopback (LU4) temperature. No other indices were statistical significant. Conclusions: Measuring the Nogung (PC9) temperature is most reasonable for assessing the severity of cold sensation of hands. To generalize this result, however, a large-scale study is warranted.

• 한국HCI학회:학술대회논문집
• /
• 2007.02a
• /
• pp.274-280
• /
• 2007

오디오 텍스처 합성은 주어진 짧은 오디오 클립으로부터 임의의 길이를 갖는 새로운 오디오 클립을 생성하는 방법이다. 이는 애니메이션이나 영화에서 비디오와 정확한 동기화를 이루는 사운드 효과를, 혹은 임의의 길이를 갖는 배경 음악을 효율적으로 만들 수 있는 방법이다. 최근 Lie Lu는 주어진 예제 오디오 클립을 여러 조각으로 나눈 후, 이 조각들을 그래프 형태로 연결하고, 생성된 그래프를 탐색하면서 임의의 길이를 가지는 오디오 클립을 합성하는 방법을 제안하였다. 비교적 간단한 방법으로도 원본 오디오 클립과 비슷한 느낌의 오디오 클립을 만들어낸다는 장점이 있지만, 이는 원본 내의 여러 오디오 조각들이 단지 지속적으로 연결되는 형태로 합성되기 때문에 종종 반복되는 느낌을 받는다는 단점이 있다. 본 논문에서는 Lie Lu의 방법과는 달리 주어진 예제 오디오 클립을 직접 합성함으로써 반복성을 줄이면서도 원본과 비슷한 느낌을 갖는 결과 오디오 클립을 생성할 수 있는 방법을 제안한다. 특히 본 논문에서는 정확한 합성을 위하여 EM 최적화 방법을 사용한다. 본 논문에서 제안하는 합성 방법은 먼저 예제 오디오 클립을 일정 단위로 나누고 이렇게 나눠진 부분들을 일정 길이만큼 서로 겹쳐지게 합성하여 임의의 길이의 오디오 클립을 만든다. 그 후 만들어진 오디오 클립을 예제 오디오 클립과 부분 부분을 비교하여 확장된 오디오 클립과 최대한 비슷한 부분을 예제 오디오 클립에서 찾는다. 그 다음 찾아진 결과를 결과 오디오에 다시 합성하여 오디오 클립을 만든다. 이런 과정을 반복하여 최적화된 가장 적절한 결과값을 구한다. 이 결과는 분할된 부분들이 가장 자연스럽게 이어지는 결과가 된다. 본 논문에서는 최적화를 사용하여 오디오를 합성하기 때문에 합성 결과를 쉽게 조정할 수 있다는 장점이 있다. 최적화 문제에 특정 제약 조건을 넣음으로써 사용자가 원하는 부분의 음악이 결과 사운드의 특정 부분에 위치 할 수 있게 하고 이로써 특정 흐름을 만들어낼 수 있으며, 일부가 손실된 사운드 데이터의 복구를 가능하게 하는 등의 결과를 생성할 수 있다. EM 최적화를 사용한 오디오 텍스처 합성 방법은 기존의 합성 방법에 비해 질적인 측면에서 보다 좋은 결과를 생성할 수 있고, 비교적 반복이 덜한 패턴들을 만들어 낼 수 있다. 이를 입증하기 위해 이에 대한 사용자 설문 조사 결과가 제시된다.