• Journal of information and communication convergence engineering
• /
• v.16 no.2
• /
• pp.106-113
• /
• 2018

Although deep learning methods of convolutional neural networks (CNNs) and long-/short-term memory (LSTM) are widely used for text categorization, they still have certain shortcomings. CNNs require that the text retain some order, that the pooling lengths be identical, and that collateral analysis is impossible; In case of LSTM, it requires the unidirectional operation and the inputs/outputs are very complex. Against these problems, we thus improved these traditional deep learning methods in the following ways: We created collateral CNNs accepting disorder and variable-length pooling, and we removed the input/output gates when creating bidirectional LSTMs. We have used four benchmark datasets for topic and sentiment classification using the new methods that we propose. The best results were obtained by combining LTSM regional embeddings with data convolution. Our method is better than all previous methods (including deep learning methods) in terms of topic and sentiment classification.

• 한국가시화정보학회:학술대회논문집
• /
• 2006.12a
• /
• pp.11-16
• /
• 2006

Reverse flow occurs in a channel when there is an obstruction at the entry, However it has been shown recently that reverse flow can be realized without an obstruction, by staggering the sides of the channel and placing it at an angle of attack to the oncoming flow. In this study the latter flow is computationally investigated. And the mechanism is investigated using PIV (Particle Image Velocimetry) method. The results have captured all the essential features of this complex phenomenon and show the time dependent pumping mechanism which leads to the occurrence of reverse flow.

• Journal of the Korean Solar Energy Society
• /
• v.31 no.2
• /
• pp.107-112
• /
• 2011

A comparison study between two performance testing results, one is on the site calibration not needed and the other is needed, was proceeded for the understanding on the effect of site calibration on the complex terrain. As a result, it is revealed that all of uncertainty components is effected by the topographical features dramatically. And the maximum difference of uncertainty reached at around 8% of rated capacity of wind turbine. So, the site calibration is an effective method to remove the variable wind effect by the ground complexity and must be proceeded before the power performance testing of a wind turbine.

• Journal of the Semiconductor & Display Technology
• /
• v.19 no.4
• /
• pp.121-125
• /
• 2020

With miniaturization of semiconductor, the manufacturing process become more complex, and undetected small changes in the state of the equipment have unexpectedly changed the process results. Fault detection classification (FDC) system that conducts more active data analysis is feasible to achieve more precise manufacturing process control with advanced machine learning method. However, applying machine learning, especially in supervised learning criteria, requires an arduous data labeling process for the construction of machine learning data. In this paper, we propose a semi-supervised learning to minimize the data labeling work for the data preprocessing. We employed equipment status variable identification (SVID) data and optical emission spectroscopy data (OES) in silicon etch with SF6/O2/Ar gas mixture, and the result shows as high as 95.2% of labeling accuracy with the suggested semi-supervised learning algorithm.

• Journal of the Korean Society of Fisheries and Ocean Technology
• /
• v.34 no.3
• /
• pp.328-338
• /
• 1998

It is main objective of this approach to present a method to analyse stochastic design sensitivity for problems of structural dynamics with randomness in design parameters. A combination of the adjoint variable approach and the second oder perturbation method is used in the finite element approach. An alternative form of the constant functional that holds for all times is introduced to consider the time response of dynamic sensitivity. The terminal problem of the adjoint system is solved using equivalent homogeneous equations excited by initial velocities. The numerical procedures are shown to be much more efficient when based on the fold superposition method : the generalized co-ordinates are normalized and the correlated random variables are transformed to uncorrelated variables, where as the secularities are eliminated by the fast Fourier transform of complex valued sequences. Numerical algorithms have been worked out and proved to be accurate and efficient : they codes whose element derivative matrices can be explicitly generated. The numerical results of two cases - 2-dimensional portal frame and 3/4-cylindrical shell structure - for the deterministic and stochastic sensitivity analysis illustrates in this paper.

• Journal of the Korean Institute of Illuminating and Electrical Installation Engineers
• /
• v.28 no.11
• /
• pp.112-122
• /
• 2014

This paper proposes a PMSM torque control method considering DC-link voltage variation and friction torque. In general EV/HEV application, two dimensions look-up table(2D-LUT) is used for reference current generation due to its stable and robust torque control performance. Conventionally, this 2D-LUT is established by flux-torque table to overcome the DC-link voltage variation. However, the flux table establishment is more complex than the speed table establishment. Moreover, one flux data reflects several speed conditions in variable DC-link voltage, friction torque cannot be considered by using the flux table. In this paper, speed-torque 2D-LUT is used for current reference generation. With this table, PMSM torque control is well achieved regardless of DC-link voltage variation by the proposed control method. Simulation and experimental results validate improvement of torque control error through friction torque compensation.

• Journal of Drive and Control
• /
• v.15 no.1
• /
• pp.28-37
• /
• 2018

When the wind speed rises above the rated wind speed, the produced power of the wind turbines exceeds the rated power. Even more, the excessive power results in the undesirable mechanical load and fatigue. A solution to this problem is pitch control of the wind turbines. This paper presents a systematic design method of a collective pitch controller for the wind turbines using a discrete fuzzy Proportional-Integral (PI) controller. Unlike conventional PI controllers, the fuzzy PI controller has variable gains according to its input variables. Generally, tuning the parameters of fuzzy PI controller is complex due to the presence of too many parameters strongly coupled. In this paper, a systematic method for the fuzzy PI controller is presented. First, we show the fact that the fuzzy PI controller is a superset of the PI controller in the discrete-time domain and the initial parameters of the fuzzy PI controller is selected by using this relationship. Second, for simplicity of the design, we use only four rules to construct nonlinear fuzzy control surface. The tuning parameters of the proposed fuzzy PI controller are also obtained by the aforementioned relationship between the PI controller and the fuzzy PI controller. As a result, unlike the PI controller, the proposed fuzzy PI controller has variable gains which allow the pitch control system to operate in broader operating regions. The effectiveness of the proposed controller is verified with computer simulations using FAST, a NREL's primary computer-aided engineering tool for horizontal axis wind turbines.

• Journal of Korean Academy of Nursing
• /
• v.54 no.2
• /
• pp.193-210
• /
• 2024

Purpose: The objective of this study was to develop a predictive model for the sexual experiences of adolescents using the random forest method and to identify the "variable importance." Methods: The study utilized data from the 2019 to 2021 Korea Youth Risk Behavior Web-based Survey, which included 86,595 man and 80,504 woman participants. The number of independent variables stood at 44. SPSS was used to conduct Rao-Scott χ² tests and complex sample t-tests. Modeling was performed using the random forest algorithm in Python. Performance evaluation of each model included assessments of precision, recall, F1-score, receiver operating characteristics curve, and area under the curve calculations derived from the confusion matrix. Results: The prevalence of sexual experiences initially decreased during the COVID-19 pandemic, but later increased. "Variable importance" for predicting sexual experiences, ranked in the top six, included week and weekday sedentary time and internet usage time, followed by ease of cigarette purchase, age at first alcohol consumption, smoking initiation, breakfast consumption, and difficulty purchasing alcohol. Conclusion: Education and support programs for promoting adolescent sexual health, based on the top-ranking important variables, should be integrated with health behavior intervention programs addressing internet usage, smoking, and alcohol consumption. We recommend active utilization of the random forest analysis method to develop high-performance predictive models for effective disease prevention, treatment, and nursing care.

• Journal of Microbiology and Biotechnology
• /
• v.29 no.4
• /
• pp.651-657
• /
• 2019

Although smallpox was eradicated in 1980, it is still considered a potential agent of biowarfare and bioterrorism. Smallpox has the potential for high mortality rates along with a major public health impact, eventually causing public panic and social disruption. Passive administration of neutralizing monoclonal antibodies (mAbs) is an effective intervention for various adverse reactions caused by vaccination and the unpredictable nature of emerging and bioterrorist-related infections. Currently, vaccinia immune globulin (VIG) is manufactured from vaccinia vaccine-boosted plasma; however, this production method is not ideal because of its limited availability, low specific activity, and risk of contamination with blood-borne infectious agents. To overcome the limitations of VIG production from human plasma, we isolated two human single-chain variable fragments (scFvs), (SC34 and SC212), bound to vaccinia virus (VACV), from a scFv phage library constructed from the B cells of VACV vaccine-boosted volunteers. The scFvs were converted to human IgG1 (VC34 and VC212). These two anti-VACV mAbs were produced in Chinese Hamster Ovary (CHO) DG44 cells. The binding affinities of VC34 and VC212 were estimated by competition ELISA to $IC_{50}$ values of $2{\mu}g/ml$ (13.33 nM) and $22{\mu}g/ml$ (146.67 nM), respectively. Only the VC212 mAb was proven to neutralize the VACV, as evidenced by the plaque reduction neutralization test (PRNT) result with a $PRNT_{50}$ of ~0.16 mg/ml (${\sim}1.07{\mu}M$). This VC212 could serve as a valuable starting material for further development of VACV-neutralizing human immunoglobulin for a prophylactic measure against post-vaccination complications and for post-exposure treatment against smallpox.

• Journal of Korea Water Resources Association
• /
• v.39 no.10 s.171
• /
• pp.867-880
• /
• 2006

Hydrological models with many parameters and complex model structures require a powerful and detailed model calibration/validation scheme. In this study, we proposed a multi-variable and multi-site calibration and validation framework for the Soil Water Assessment Tool (SWAT) model applied in the Gap-cheon catchment located downstream of the Geum river basin. The sensitivity analysis conducted before main calibration helped understand various hydrological processes and the characteristics of subcatchments by identifying sensitive parameters in the model. In addition, the model's parameters were estimated based on existing data prior to calibration in order to increase the validity of model. The Nash-Sutcliffe coefficients and correlation coefficient were used to estimate compare model output with the observed streamflow data: $R_{eff}\;and\;R^2$ ranged 0.41-0.84 and 0.5-0.86, respectively, at the Heuduck station. Model reproduced baseflow estimated using recursive digital filter except for 2-5% overestimation at the Sindae and Boksu stations. Model also reproduced the temporal variability and fluctuation magnitude of observed groundwater levels with $R^2$ of 0.71 except for certain periods. Therefore, it was concluded that the use of multi-variable and multi-site method provided high confidence for the structure and estimated parameter values of the model.