• Journal of the Korean Operations Research and Management Science Society
• /
• v.38 no.1
• /
• pp.139-152
• /
• 2013

The purpose of this study is to claim the validity of tuning the architecture of neural network models for multi-class classification. A neural network model for multi-class classification is basically constructed by building a series of neural network models for binary classification. Building a neural network model, we are required to set the values of parameters such as number of hidden nodes and weight decay parameter in advance, which draws special attention as the performance of the model can be quite different by the values of the parameters. For better performance of the model, it is absolutely necessary to have a prior process of tuning the parameters every time the neural network model is built. Nonetheless, previous studies have not mentioned the necessity of the tuning process or proved its validity. In this study, we claim that we should tune the parameters every time we build the neural network model for multi-class classification. Through empirical analysis using wine data, we show that the performance of the model with the tuned parameters is superior to those of untuned models.

• Geomechanics and Engineering
• /
• v.10 no.1
• /
• pp.37-48
• /
• 2016

The hyperbolic stress-strain model has been shown to be valid for modeling nonlinear stress-strain behavior for rockfill materials. The Duncan-Chang nonlinear constitutive model was adopted to characterize the behavior of the modeled rockfill materials in this study. Accurately estimating the model parameters of rockfill materials is a key problem for simulating dam deformations during both the dam construction period and the dam operation period. In order to estimate model parameters, triaxial compression experiments of rockfill materials were performed. Based on a genetic algorithm, the constitutive model parameters of the rockfill material were determined from the triaxial compression experimental data. The investigation results show that the predicted strains provide satisfactory precision when compared with the observed strains and the strains forecasted by a gradient-based optimization algorithm. The effectiveness of the proposed inversion procedure of model parameters was verified by experimental investigation in a laboratory.

• Journal of the Earthquake Engineering Society of Korea
• /
• v.27 no.6
• /
• pp.275-281
• /
• 2023

The demand for high-strength steel is rising due to its economic efficiency. Low-cycle fatigue (LCF) tests have been conducted to investigate the nonlinear behaviors of high-strength steel. Accurate material models must be used to obtain reliable results on seismic performance evaluation using numerical analyses. This study uses the combined hardening model to simulate the LCF behavior of high-strength steel. However, it is challenging and complex to determine material model parameters for specific high-strength steel because a highly nonlinear equation is used in the model, and several parameters need to be resolved. This study used the particle swarm algorithm (PSO) to determine the model parameters based on the LCF test data of HSA 650 steel. It is shown that the model with parameter values selected from the PSO accurately simulates the measured LCF curves.

• Magazine of the Korean Society of Agricultural Engineers
• /
• v.40 no.1
• /
• pp.57-67
• /
• 1998

A stochastic weather generator which simulate daily precipitation, maximum and minimum daily temperature, relative humidity was developed. The model parameters were estimated using stochastic characteristics analysis of historical data of 71 weather stations. Spatial variations of the parameters for the country were also analyzed. Model parameters of ungauged Sites were determined from parameters of adjacent weather stations using inverse distance method. The model was verified on Suwon and Ulsan weather stations and showed good agreement between simulated and observed data.

• The Bulletin of The Korean Astronomical Society
• /
• v.46 no.2
• /
• pp.46.3-47
• /
• 2021

Recently a multilayer spectral inversion (MLSI) model has been proposed to infer the physical parameters of plasmas in the solar chromosphere. The inversion solves a three-layer radiative transfer model using the strong absorption line profiles, H alpha and Ca II 8542 Å, taken by the Fast Imaging Solar Spectrograph (FISS). The model successfully provides the physical plasma parameters, such as source functions, Doppler velocities, and Doppler widths in the layers of the photosphere to the chromosphere. However, it is quite expensive to apply the MLSI to a huge number of line profiles. For example, the calculating time is an hour to several hours depending on the size of the scan raster. We apply deep neural network (DNN) to the inversion code to reduce the cost of calculating the physical parameters. We train the models using pairs of absorption line profiles from FISS and their 13 physical parameters (source functions, Doppler velocities, Doppler widths in the chromosphere, and the pre-determined parameters for the photosphere) calculated from the spectral inversion code for 49 scan rasters (~2,000,000 dataset) including quiet and active regions. We use fully connected dense layers for training the model. In addition, we utilize a skip connection to avoid a problem of vanishing gradients. We evaluate the model by comparing the pairs of absorption line profiles and their inverted physical parameters from other quiet and active regions. Our result shows that the deep learning model successfully reproduces physical parameter maps of a scan raster observation per second within 15% of mean absolute percentage error and the mean squared error of 0.3 to 0.003 depending on the parameters. Taking this advantage of high performance of the deep learning model, we plan to provide the physical parameter maps from the FISS observations to understand the chromospheric plasma conditions in various solar features.

• 제어로봇시스템학회:학술대회논문집
• /
• 1999.10a
• /
• pp.153-156
• /
• 1999

The technique is reported of identifying parameters in off-line process. The technique demands that closed-loop system consists of a reference and two-degree-of-freedom controllers (TDFC) in real process. A model process is the same as the real process except their parameters. Deviations are differences between the reference and the output of the plant or the model. The technique is based on minimizing identification error between the two deviations. The parameter differences between the plant and the model are characterized of mean value and of variance which are derived from the identification error. Consequently, the algorithm which identifies the unknown plant parameters is shown by minimizing the mean value and the variance, respectively, within double convergence loops. The technique is applied to course change of a ship. The plant deviation at the first trial is shown to occur in replacing the nominal parameters by the default parameters. The plant deviation at the second trial is shown to not occur in replacing the nominal parameters by the identified parameters. Hence, the identification technique is confirmed to be feasible in the real field.

• Phonetics and Speech Sciences
• /
• v.9 no.4
• /
• pp.91-97
• /
• 2017

This study categorizes several voice characteristics by subjective listening assessment, and investigates correlation between voice characteristics and speech feature parameters. A model was developed to classify voice characteristics into the defined categories using SVM algorithm. To do this, we extracted various speech feature parameters from speech database for men in their 20s, and derived statistically significant parameters correlated with voice characteristics through ANOVA analysis. Then, these derived parameters were applied to the proposed SVM model. The experimental results showed that it is possible to obtain some speech feature parameters significantly correlated with the voice characteristics, and that the proposed model achieves the classification accuracies of 88.5% on average.

• Journal of the Korean Institute of Electrical and Electronic Material Engineers
• /
• v.11 no.5
• /
• pp.347-352
• /
• 1998

In this paper, the extraction and optimization of SPICE parameters on SONOSFET for NVSM circuit design were discussed. SONOSFET devices with different channel widths and lengths were fabricated using conventional 1.2 um n-well CMOS process. And, electric properties for dc parameters and capacitance parameters were measured on wafer. SPICE parameters for the SONOSFET were extracted from the UC Berkeley level 3 model for the MOSFET. And, local optimization of Ids-Vgs curves has carried out in the bias region of subthreshold, linear, saturation respectively. Finally, the extracted SPICE parameters were optimized globally by comparing drain current (Ids), output conductance(gds), transconductance(gm) curves with theoretical curves in whole region of bias conditions. It is shown that the conventional model for the MOSFET can be applied to the SONOSFET modeling except sidewalk effect.

• Proceedings of the Computational Structural Engineering Institute Conference
• /
• 2008.04a
• /
• pp.96-99
• /
• 2008

When the design modification is occurred, at present, design process based on 2-D spend more time and effort than that based on 3-D to modify related structural details. To improve and develop these processes, therefore, the design possibility of civil structures based on virtual model of 3-D must be investigated. We designed reinforced concrete pier of 3-D model, containing parameters. The parameters was defined as structural details like area of the section, reinforcement specification for design modification and structural analysis. In this paper, we researched about the processes modeling of reinforced concrete bridge pier based on parameters, the extracting data from the virtual model of 3-D, and the reflection of data to virtual model throughout structural analysis.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2009.05a
• /
• pp.107-111
• /
• 2009

The parameters in the Horton's model which has well known as typical infiltration model were determined by the use of the optimization technique. It was assumed the initial infiltration capacity in this model was related to the antecedent precipitation per 10 days with linear combination. And both the parameters of the ultimate infiltration capacity and the decay factor were determined uniquely on a basin. Thus the optimal model's parameters representative to a basin were obtained and the Horton's infiltration equations by rainstorm events were determined. The data of ten rainstorm events for this study were observed at the Jeonjeokbigyo station located at the Selmacheon experimental basin that was $8.5\;km^2$ wide in the Imjin river.