• Journal of IKEEE
• /
• v.22 no.1
• /
• pp.14-20
• /
• 2018

The purpose of training a convolutional neural network (CNN) is to obtain weight factors that give high classification accuracies. The initial values of hyper-parameters affect the training results, and it is important to train a CNN with a suitable hyper-parameter set of a learning rate, a batch size, the initialization of weight factors, and an optimizer. We investigate the effects of a single hyper-parameter while others are fixed in order to obtain a hyper-parameter set that gives higher classification accuracies and requires shorter training time using a proposed VGG-like CNN for training since the VGG is widely used. The CNN is trained for four datasets of CIFAR10, CIFAR100, GTSRB and DSDL-DB. The effects of the normalization and the data transformation for datasets are also investigated, and a training scheme using merged datasets is proposed.

• Proceedings of the KSME Conference
• /
• 2000.04a
• /
• pp.194-199
• /
• 2000

The creep rate is affected by the temperature and in fact. if the temperature above $T_M/2(T_M:melting\;point)$. The aim of the present investigation is to study the relationship of static creep and cyclic creep behavior of pure copper and the formulation of these phenomena with the special attention to the instantaneous strain. strain rate from time and number of cycles have the same inclination Steady state creep rate depend upon maximum stress and can be expressed as linear function according to Power law creep equations Creep rupture time has relation with creep rate. and it make a group represented as the same direct line regardless of max stress, stress ratio and the temperature. Initial strain effect on continuous creep deformation. and have guantitative relationship between elastic and Plastic strain. LMP have similar tendency than OSDP and MHP according to temperature

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2009.01a
• /
• pp.408-412
• /
• 2009

In this paper, a new efficient algorithm for global motion estimation is proposed. This algorithm uses a previous 4-parameter model based global motion estimation algorithm and M-estimator for improving the accuracy and robustness of the estimate. The first algorithm uses the block based motion vector fields and which generates a coarse global motion parameters. And second algorithm is M-estimator technique for getting precise global motion parameters. This technique does not increase the computational complexity significantly, while providing good results in terms of estimation accuracy. In this work, an initial estimation for the global motion parameters is obtained using simple 4-parameter global motion estimation approach. The parameters are then refined using M-estimator technique. This combined algorithm shows significant reduction in mean compensation error and shows performance improvement over simple 4-parameter global motion estimation approach.

• Journal of Korea Water Resources Association
• /
• v.33 no.S1
• /
• pp.3-10
• /
• 2000

A newly developed heuristic algorithm, Harmony Search, is applied to the parameter calibration problem of the nonlinear Muskingum model. The Harmony Search could, mimicking the improvisation of music player, find better parameter values for in the nonlinear Muskingum model than five other methods including another heuristic method, genetic algorithm, in the aspect of SSQ(the sum of the square of the deviations between the observed and routed outflows) as well as in the aspects of SAD(the sum of the absolute value of the deviations), DPO(deviations of peak of routed and actual flows) and DPOT(deviatios of peak time of routed and actual outflow). Harmony Search also has the advantage that it does not require the process of asuming the initial values of desing parameters. The sensitivity analysis of Harmony Memory Considering Rate showed that relatively large values of Harmony Memory Considering Rate makes the Harmony Search converge to a better solution.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.22 no.1
• /
• pp.185-192
• /
• 1997

We propose a method of paramete estimation using order-of-magnitude analysis for optimal boundary smoothing in Mean Field Annealing(MFA) technique in this paper. We previously proposed two boundary smoothing methods for consistent object representation in the previous paper, one is using a constratined regulaization(CR) method and the other is using a MFA method. The CR method causes unnecessary smoothing effects at corners. On the other hand, the MFA method method smooths our the noise without losing sharpness of corners. The MFA algorithm is influenced by several parameters such as standard deviation of the noise, the relativemagnitude of prior ter, initial temperature and final temperature. We propose a general parameter esimation method for optimal boundary smoothing using order-of-magnitude analysis to be used for consistent object representation in this paper. In addition, we prove the effectiveness of our parameter estimation and also show the temperature parameter sensitivities of the algorithm.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.5 no.4
• /
• pp.297-301
• /
• 2005

In this paper, we propose a method to estimate camera motion parameter based on invariant point features. Typically, feature information of image has drawbacks, it is variable to camera viewpoint, and therefore information quantity increases after time. The LM(Levenberg-Marquardt) method using nonlinear minimum square evaluation for camera extrinsic parameter estimation also has a weak point, which has different iteration number for approaching the minimal point according to the initial values and convergence time increases if the process run into a local minimum. In order to complement these shortfalls, we, first propose constructing feature models using invariant vector of geometry. Secondly, we propose a two-stage calculation method to improve accuracy and convergence by using homography and LM method. In the experiment, we compare and analyze the proposed method with existing method to demonstrate the superiority of the proposed algorithms.

• The Transactions of the Korean Institute of Power Electronics
• /
• v.25 no.3
• /
• pp.227-234
• /
• 2020

This paper presents the methods for the modeling and parameter optimization of the electric vehicle battery. The state variables of the battery are defined, and the test methods for battery parameters are presented. The state-space equation, which consists of four state variables, and the output equation, which is a combination of to-be-determined parameters, are shown. The parameter optimization method is the key point of this study. The least square of the modeling error can be used as an initial value of the multivariable function. It is equivalent to find the minimum value of the error function to obtain optimal parameters from multivariable function. The SIMULINK model is presented, and the 10-hour full operational range test results are shown to verify the performance of the model. The modeling error for 25 degrees is approximately 1% for full operational ranges. The comments to enhance modeling accuracy are shown in the conclusion.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2000.05a
• /
• pp.3-10
• /
• 2000

A newly developed heuristic algorithm, Harmony Search, is applied to the parameter calibration problem of the nonlinear Muskingum model. The Harmony Search could, mimicking the improvisation of music players, find better parameter values for in the nonlinear Muskingum model than five other methods including another heuristic method, genetic algorithm, in the aspect of SSQ (the sum of the square of the deviations between the observed and routed outflows) as well as in the aspects of SAD (the sum of the absolute value of the deviations), DPO (deviations of peak of routed and actual flows) and DPOT (deviations of peak time of rented and actual outflow). Harmony Search also has the advantage that it does not require the process of assuming the initial values of design parameters. The sensitivity analysis of Harmony Memory Considering Rate showed that relatively large values of Harmony Memory Considering Rate makes the Harmony Search converse to a better solution.

• The Transactions of the Korean Institute of Power Electronics
• /
• v.17 no.3
• /
• pp.238-245
• /
• 2012

This paper presents a sensorless speed control of a 2-phase switch reluctance motor(SRM). The proposed sensorless control scheme is based on the slide mode observer with parameter compensator to improve the estimation performance. In the stand still position, the initial rotor position is determined by pulse current responses of each phase windings and the current difference. In order to determine an accurate initial rotor position, the two initial rotor positions are estimated by the difference of the pulse currents. From the stand still to the operating region, a simple open loop control which determines the commutation sequence by the pulse current of the unexcited phase winding is used. When the motor speed is reached to the sensorless control region, the estimated rotor position and speed by the slide mode observer are used to control the SRM. The flux calculator used in the slide mode observer is designed by phase voltage and the voltage drops in the phase resistance of the winding. The accuracy of the flux calculator is dependent on the phase resistance. For the continuous update of the phase resistance, current gradient at the inductance break point is used in this paper. The error of the estimated rotor position at the current gradient position is used to update the phase resistance to improve the sensorless scheme. The proposed sensorless speed control scheme is verified with a practical compressor used in home appliances. And the results show the effectiveness of the proposed control scheme.

• Applied Chemistry for Engineering
• /
• v.18 no.5
• /
• pp.424-432
• /
• 2007

With the aiming to improve the comprehension of an expansion process of perlite, a numerical study has been carried out, concerning a water vapor bubble growth in a softened perlite melt. Isothermal case of the model is investigated and compared with the experimental results. Water content is an intrinsic parameter in a perlite expansion. With increasing water content, the expansion time is faster. But it can not be changed easily in an industrial process. Initial temperature is a changeable parameter in an industrial process. Higher initial temperature hastens the expansion time and this can be a key parameter in an industrial process. Experimental results are observed by SEM. Increasing initial temperature, perlite shape becomes close to sphere. Calculated bubble growth results are compared with experimental results. Packing density is measured and compared with simulated results. These results indicate that some other factors are applied on the model during heat transformation to work.