• Journal of Advanced Information Technology and Convergence
• /
• v.9 no.1
• /
• pp.127-134
• /
• 2019

In terms of deep Reinforcement Learning (RL), exploration can be worked stochastically in the action of a state space. On the other hands, exploitation can be done the proportion of well generalization behaviors. The balance of exploration and exploitation is extremely important for better results. The randomly selected action with ε-greedy for exploration has been regarded as a de facto method. There is an alternative method to add noise parameters into a neural network for richer exploration. However, it is not easy to predict or detect over-fitting with the stochastically exploration in the perturbed neural network. Moreover, the well-trained agents in RL do not necessarily prevent or detect over-fitting in the neural network. Therefore, we suggest a novel design of a deep RL by the balance of the exploration with drop-out to reduce over-fitting in the perturbed neural networks.

• The Journal of Information Systems
• /
• v.31 no.1
• /
• pp.25-58
• /
• 2022

Purpose The purpose of this study is to analyzes the RPA fitting processes by the casese of manufacturing SMEs(Small and Medium-sized Enterprises) in an exploraty approach. Based on the findings on the RPA fitting processes, we intend to provide a cornerstone for developing a general-purpose RPA introduction model in the future. Design/methodology/approach In this study, empirical cases of RPA fitting processes were analyzed based on interviews with project managers of specialized IT suppliers in charge of RPA development and managers of IT departments of manufacturing SMEs that actually introduced RPA. In order to explore various RPA fitting process in the manufacturing value chain, a total of 7 manufacturing SMEs were interviewed, ranging from companies using a legacy system to companies without a legacy system. Over the primary and secondary activity processes, the details of RPA processes were analyzed in the steps of 'Frequency Identification, Input Processing, Source Identification, Inquiry and Processing, Information Registration, Result Reporting'. Findings From the analysis, we derived some exploratory results that the processes over 0.25 FTE and related with many suppliers and clients are fitting for RPA introduction in manufacturing SMEs Our results will provide basic data for the development of the future general-purpose RPA introduction model for manufacturing SMEs, providing practical reference for RPA introduction.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.57 no.11
• /
• pp.2095-2101
• /
• 2008

In this paper, we consider a performance improvement of neural network for predicting defect size of steam generator tube using early stopping. Usually, neural network is trained until MSE becomes less than a prescribed error goal. The smaller the error goal, the greater the prediction performance for the trained data. However, as the error goal is decreased, an over fitting is likely to start during supervised training of a neural network, which usually deteriorates the generalization performance. We propose that, for the prediction of an axisymmetric defect size, early stopping can be used to avoid the over-fitting. Through various experiments on the axisymmetric defect samples, we found that the difference bet ween the prediction error of neural network based on early stopping and that of ideal neural network is reasonably small. This indicates that the error goal used for neural network training for the prediction of defect size can be efficiently selected by early stopping.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.30 no.10 s.253
• /
• pp.1219-1226
• /
• 2006

The ASTM test standard recommends the use of the compact tension specimen for creep crack growth rates measurement. In the creep crack growth rate test, the displacement rate due to creep is obtained by subtracting the contribution of elastic and plastic components from the total load line displacement rate based on displacement partitioning method fur determining $C^*-integral$, which involves Ramberg-Osgood (R-O) fitting procedures. This paper investigates the effect of the R-O fitting procedures on plastic displacement rate estimates in creep crack growth testing, via detailed two-dimensional and three-dimensional finite element analyses of the standard compact tension specimen. Four different R-O fitting procedures are considered; (i) fitting the entire true stress-strain data up to the ultimate tensile strength, (ii) fitting the true stress-strain data from 0.1% strain to 0.8 of the true ultimate strain, (iii) fitting the true stress-strain data only up to 5% strain, and (iv) fitting the engineering stress-strain data. It is found that the last two procedures provide reasonably accurate plastic displacement rates and thus should be recommended in creep crack growth testing. Moreover, several advantages of fitting the engineering stress-strain data over fitting the true stress-strain data only up to 5% strain are discussed.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.23 no.4
• /
• pp.388-393
• /
• 2019

As a fiiting method to data with outliers, RANSAC(RANdom SAmple Consensus) based algorithm is widely used in fitting of line, circle, ellipse, etc. CUDA is currently most widely used GPU with massive parallel processing capability. This paper proposes an efficient CUDA implementation of multiple planes fitting using RANSAC with 3d points data, of which one set of 3d points is used for one plane fitting. The performance of the proposed algorithm is demonstrated compared with CPU implementation using both artificially generated data and real 3d heights data of a PCB. The speed-up of the algorithm over CPU seems to be higher in data with lower inlier ratio, more planes to fit, and more points per plane fitting. This method can be easily applied to a wide variety of other fitting applications.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.62 no.7
• /
• pp.913-917
• /
• 2013

During the restoration process of primary restorative transmission system, some over voltages may happen due to nonlinear interaction between unloaded transformers and transmission systems. These over voltages caused by harmonic resonance can be suppressed by inserting damping loads before energizing transformers. But it is very difficult to predict the occurrence possibility of harmonic resonance and complex simulation must be repeated to estimate the sufficient damping loads. This paper presents a damping loads prediction system to prevent harmonic resonance. Detailed analysis of the relationship between harmonic resonance and the amount of damping loads is discussed. The prediction system is developed using a curve fitting and a neural network based on this relationship. A curve fitting used a Gaussian function based on non-linear least square method and multi-layer back-propagation neural network is applied. The system is applied to primary restorative transmission lines in korean power system and the result showed satisfactory performance.

• 제어로봇시스템학회:학술대회논문집
• /
• 1991.10a
• /
• pp.917-921
• /
• 1991

Hydraulic fitting impulse tester is equipment which produce impulse pressure waveform that specified foreign standard of SAE, JIS etc. Test conditions of SAE J1453 about waveform standard indicates frequency of 35-70 cycle/min, pressure of 560 bar, oil temperature of 93 .+-.3.deg. C etc. and required cycle is a million over. In additions, Test condition operated continuously equipment. This development item adopted new pattern method such as intensifier and rotary distributor is different from already established fitting impulse tester applied servo valve and high pressure direct directional control valve. Therefore, this development item which compares already established item is good reliability, low cost of manufacture and save of electric energy. especially, Domestic small and medium enterprise uses this tester because of economical cause. We develope appropriateness tester which conforms to demand of user.

• Proceedings of the Korean Institute of IIIuminating and Electrical Installation Engineers Conference
• /
• 2008.05a
• /
• pp.383-386
• /
• 2008

We developed a fiber-optic Bragg grating sensor system using a SOA fiber laser for over heat detection in power systems. To compensate the nonlinear wavelength tuning of the fiber laser, we used fixed passband wavelengths from Fabry-Perot ITU filter as reference wavelengths. Gaussian line-fitting algorithm was also used to reduce the FBG peak detection error. Compared with a highest-peak-detection and a polynomial-fitting method, the proposed Gaussian fitting algorithm could drastically reduce the measurement errors. Also the SOA fiber laser made it possible to enhance the signal-to-noise-ratio even with several kilometers of lead fiber.

• Proceedings of the Korean Information Science Society Conference
• /
• 2007.10c
• /
• pp.409-414
• /
• 2007

Active Appearance Model is a well-known model that can represent a non-rigid object effectively. However, since it uses the fixed appearance model, the fitting results are often unsatisfactory when the imaging condition of the target image is different from that of training images. To alleviate this problem, incremental AAM was proposed which updates its appearance bases in an on-line manner. However, it cannot deal with the sudden changes of illumination. To overcome this, we propose a novel scheme to update the appearance bases. When a new person appears in the input image, we synthesize illuminated images of that person and update the appearance bases of AAM using it. Since we update the appearance bases using synthesized illuminated images in advance, the AAM can fit their model to a target image well when the illumination changes drastically. The experimental results show that our proposed algorithm improves the fitting performance over both the incremental AAM and the original AAM.

• Ocean Systems Engineering
• /
• v.4 no.4
• /
• pp.293-308
• /
• 2014

Unlike the traditional displacement type vessels, the high speed planing crafts are supported by the lift forces which are highly non-linear. This non-linear phenomenon causes their motions in an irregular seaway to be non-Gaussian. In general, it may not be possible to express the probability distribution of such processes by an analytical formula. Also the process might not be stationary or ergodic in which case the statistical behavior of the motion to be constantly changing with time. Therefore the extreme values of such a process can no longer be calculated using the analytical formulae applicable to Gaussian processes. Since closed form analytical solutions do not exist, recourse is taken to fitting a distribution to the data and estimating the statistical properties of the process from this fitted probability distribution. The peaks over threshold analysis and fitting of the Generalized Pareto Distribution are explored in this paper as an alternative to Weibull, Generalized Gamma and Rayleigh distributions in predicting the short term extreme value of a random process.