• Journal of Information Processing Systems
• /
• v.14 no.5
• /
• pp.1075-1086
• /
• 2018

In this paper a cross-validation algorithm for training probabilistic neural networks (PNNs) is presented in order to be applied to automatic face identification. Actually, standard PNNs perform pretty well for small and medium sized databases but they suffer from serious problems when it comes to using them with large databases like those encountered in biometrics applications. To address this issue, we proposed in this work a new training algorithm for PNNs to reduce the hidden layer's size and avoid over-fitting at the same time. The proposed training algorithm generates networks with a smaller hidden layer which contains only representative examples in the training data set. Moreover, adding new classes or samples after training does not require retraining, which is one of the main characteristics of this solution. Results presented in this work show a great improvement both in the processing speed and generalization of the proposed classifier. This improvement is mainly caused by reducing significantly the size of the hidden layer.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 1994.10a
• /
• pp.719-724
• /
• 1994

This paper describes a hybrid surface-based method for smooth 3D surface approximation to a sequence of 2D cross sections. The resulting surface is a hybrid G $^{1}$ surface represented by a mesh of triangular and rectangular Bezier patches defined on skinning, branching, or capping regions. Each skinning region is approximated with a closed B_spline surface, which is transformed into a mesh of Bezier patches. Triangular G $^{1}$ surfaces are constructed over brabching and capping regions such that the transitions between each capping regions such that the transitions between each triangular surface and its neighboring skinning surfaces are G $^{1}$ continuous. Since each skinning region is represented by an approximated rectangular C $^{2}$ suface instead of an interpolated trctangular G $^{[-1000]}$ surface, the proposed method can provide more smooth surfaces and realize more efficient data reduction than triangular surfacebased method.

• Analytical Science and Technology
• /
• v.8 no.4
• /
• pp.663-668
• /
• 1995

Ion transport of a polypyrrole/chloride (PPy/Cl) film and a polypyrrole/poly(styenesulfonate) (PPy/PSS) film as a function of applied dc potential was investigated by employing electrogravimetric impedance technique and electrochemical impedance technique. The cation and anion contribution to the whole charge capacitance and the diffusion coefficients of cation and anion in a PPy/PSS film were calculated by fitting the electrogravimetric impedance data with proposed model circuit. The diffusion coefficients of $Na^+$ in a 1 M $NaClO_4$ solution are over 1 order of magnitude larger than those of $ClO{_4}^-$, and $ClO{_4}^-$ contribution to charge compensation decreases as dc potential lowers. The charge compensation of a PPy/Cl film ir a 1 M CsCl solution is carried out largely by $Cl^-$ at 0.2 V vs. Ag/AgCl and by $Cs^+$ as well as $Cl^-$ at -0.4 V.

• International Journal of Aeronautical and Space Sciences
• /
• v.17 no.2
• /
• pp.149-156
• /
• 2016

The hypersonic flow on the stagnation streamline of a blunt body is analyzed with quasi one-dimensional (1-D) Navier-Stokes equations approximated by adopting the local similarity to the two-dimensional (2-D)/axisymmetric Navier-Stokes equations. The governing equations are solved using the implicit finite volume method. The computational domain is confined from the stagnation point to the shock wave, and the shock fitting method is used to find the shock position. We propose a boundary condition at the shock, which employs the shock wave angle in the vicinity of the stagnation streamline using the shock shape correlation. As a result of numerical computation conducted for the hypersonic flow over a sphere, the proposed boundary condition is shown to improve the accuracy of the prediction of the shock standoff distance. The quasi 1-D Navier-Stokes code is efficient in computing time and is reliable for the flow analysis along the stagnation streamline and the prediction of heat flux at the stagnation point in the hypersonic blunt body flow.

• Journal of Veterinary Clinics
• /
• v.20 no.3
• /
• pp.364-368
• /
• 2003

Cytogenetic and hematological analysis was performed in peripheral blood from the cattle associated with abnormal delivery around nuclear power plant area. The frequency of micronuclei (MN) in peripheral blood lymphocytes from cattle was used as a biomarker of radiobiological effects resulting from exposure to environmental radiation. An estimated dose of radiation was calculated by best fitting linear-quadratic model based on the radiation-induced MN data over the range from 0 Gy to 4 Gy from the bovine lymphocytes with in vitro irradiation. MN rates in live malformed calf, dams of malformed calves and other cattle living in the same barn from the regions around nuclear power plant, and cattle in control area were 9/1000, 10.8/1000, 13.3/1000 and 10.0/1000, respectively. There were no significant differences in MN frequencies and hematological values between the cattle associated with abnormal delivery around nuclear power plant area and those of control area. This study indicates that the congenital abnormalities near nuclear power plant seemed to be caused by other aetiology.

• The International Journal of Costume Culture
• /
• v.2 no.2
• /
• pp.1-13
• /
• 1999

This study aimed to give an updated description of general clothing behavior patterns of the Korean elderly women. 430 women over 60 years old residing in Seoul wre surveyed with a questionnaire asking about their daily practices of clothing selection, disposal, recycling, quantity and place of purchase, and their background variables. Data were treated to SPSS for descriptive statistics, Chi-square analysis, ANOVA and Duncan's multiple range test. The result revealed that the wardrobe of the Korean aged women is by now fully oriented toward the western style. The subjects disposed of their clothing mainly because of their body size changes and had elementary fitting problems unsolved. They had clothing recycling experinces and turned out to be inactive clothing consumers with differing shopping places according to the types of clothing items to be purchased. The subjects'age and income level had significant effects on various facets of their clothing behavior.

• Korean Journal of Veterinary Research
• /
• v.44 no.3
• /
• pp.343-348
• /
• 2004

Cytogenetic and hematological analysis was performed in peripheral blood of cattle in the vicinity of Uljin nuclear power station and control area. The frequency of micronuclei(MN) in peripheral blood lymphocytes from cattle was used as a biomarker of radiobiological effects resulting from exposure to environmental radiation. An estimated dose of radiation was calculated by a best fitting linear-quadratic model based on the radiation-induced MN formation from the bovine lymphocytes exposed in vitro to radiation over the range from 0 Gy to 4 Gy. MN ratio in lymphocytes of cattle from Uljin nuclear power station and control area were 8.90/1,000 and 9.60/1,000, respectively. There were no significant differences in MN frequencies and hematological values in cattle between Uljin and control area.

• The Bulletin of The Korean Astronomical Society
• /
• v.41 no.1
• /
• pp.72.3-72.3
• /
• 2016

A massive envelope and a strong bipolar outflow are the two most distinct structures of youngest protostellar systems. We present observational results from the Combined Array for Research in Millimeter-wave Astronomy (CARMA) toward the youngest (Class 0) protostellar system L1157. At an angular resolution of 5 arcseconds, we mapped its well-developed outflow in CO 2-1 over a span of approximately 5 arcminutes. Additionally, we imaged the central envelope with CO isotopes, CS, CN, and N2H+ with an angular resolution of about 2 arcseconds. We show that the bipolar outflow may be represented with a two jet model and constrain its physical properties such as precession/rotation directions, velocities, inclinations, and position angles via cube data fitting. In addition, we discuss the kinematic features of the envelope detected in CO isotopes and N2H+ and present the radius-dependent dust opacity spectral index.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.14 no.3
• /
• pp.209-215
• /
• 2014

Sequence tagging is the task of predicting frame-wise labels for a given input sequence and has important applications to diverse domains. Conventional methods such as maximum likelihood (ML) learning matches global features in empirical and model distributions, rather than local features, which directly translates into frame-wise prediction errors. Recent probabilistic sequence models such as conditional random fields (CRFs) have achieved great success in a variety of situations. In this paper, we introduce a novel discriminative CRF learning algorithm to minimize local feature mismatches. Unlike overall data fitting originating from global feature matching in ML learning, our approach reduces the total error over all frames in a sequence. We also provide an efficient gradient-based learning method via gradient forward-backward recursion, which requires the same computational complexity as ML learning. For several real-world sequence tagging problems, we empirically demonstrate that the proposed learning algorithm achieves significantly more accurate prediction performance than standard estimators.

• Proceedings of the Korean Institute of Resources Recycling Conference
• /
• 2006.09a
• /
• pp.3-12
• /
• 2006

The main feature of these total technologies is that we can constitute the optimum treatment scheme fitting to the property of wastes, amount of wastes and energy requirement. For high moisture content wastes or biomass resources, high pressure steam process (MMCS) for crush, dry and deodorize wastes to produce high quality fertilizer of fuel is most appropriate. For dry or semi-dry solid wastes, the STAR-MEET system can be applied to produce low-BTU gases for power generation using duel fueled diesel engines of Stirling engines, and the REPRES and HyPR-MEET systems can be applied to produce hydrogen rich medium-BTU gas. For waste plastics and oils, liquefaction technology is best fit to produce light oil or kerosene equivalent fuel oils. These total technologies are completely different from the existent waste treatment technologies based on land-filling or incineration, and are expected to disseminate all over the world in the near future.