• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.33B no.9
• /
• pp.49-61
• /
• 1996

Back-propagation(BP) algorithm needs a lot of time to train the artificial neural network (ANN) to get high accuracy level in classification tasks. So there have been extensive researches to process back-propagation algorithm on parallel processors. This paper prsents a linear systolic array which calculates forward-backward propagation of BP algorithm at the same time using effective space-time transformation and PE structure. First, we analyze data flow of forwared and backward propagations and then, represent the BP algorithm into data dapendency graph (DG) which shows parallelism inherent in the BP algorithm. Next, apply space-time transformation on the DG of ANN is turn with orthogonal direction projection. By doing so, we can get a snakelike systolic array. Also we calculate the interval of input for parallel processing, calculate the indices to make the right datas be used at the right PE when forward and bvackward propagations are processed in the same PE. And then verify the correctness of output when forward and backward propagations are executed at the same time. By doing so, the proposed system maximizes parallelism of BP algorithm, minimizes th enumber of PEs. And it reduces the execution time by 2 times through making idle PEs participate in forward-backward propagation at the same time.

• Journal of the Korean Data and Information Science Society
• /
• v.16 no.3
• /
• pp.567-580
• /
• 2005

It is usually assumed that asset returns in the stock market are normally distributed. However, analyses of real data show that the distribution tends to be skewed and to have heavier tails than those of the normal distribution. In this paper, we investigate the method of estimating the value at risk(VaR) of stock returns. The VaR is computed by using the transformation and back-transformation method. The analysis of KOSPI and KOSDAQ data shows that the proposed estimation outperformed that under the normal assumption.

• Transactions of Materials Processing
• /
• v.19 no.8
• /
• pp.468-473
• /
• 2010

Shape memory alloys (SMAs) have the ability to recover their original shape upon thermo-mechanical loading even after large inelastic deformation. The unique feature is known as pseudoelasticity and shape memory effect caused by the crystalline structural transformation between two solid-state phases called austenite and martensite. To support the engineering application, a number of constitutive models, which can be formally classified into either micromechanics-based or phenomenological model, have been developed. Most of the constitutive models include a kinetic law governing the crystallographic transformation. The present work presents a one-dimensional, phenomenological constitutive model for SMAs in the context of the unified viscoplasticity theory. The proposed model does not incorporate the complex mechanisms of phase transformation. Instead, the effects induced by the transformation are depicted through the growth law for the back stress that is an internal state variable of the model.

• The Korean Journal of Applied Statistics
• /
• v.33 no.4
• /
• pp.357-364
• /
• 2020

We review the characteristics of a geometric mean and statistical inferences based on geometric means. We also show that the statistical results obtained by the logarithmic transform and back-transformation are related to geometric means and explain how to interpret the results produced in this process.

• The Korean Journal of Applied Statistics
• /
• v.22 no.5
• /
• pp.971-979
• /
• 2009

In this paper, we introduce the method that reduces the bias when the transformation and back-transformation approach is applied in GARCH models. A parametric bootstrap is employed to compute the conditional expectation which is the prediction value to minimize mean square errors in the original scale. Through the analyese of returns of KOSPI and KOSDAQ, we verified that the proposed method provides a bias-reduced estimation for the prediction value.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.32B no.5
• /
• pp.765-776
• /
• 1995

This paper describes a speaker adaptation method to improve the recognition performance of MLP(multiLayer Perceptron) based HMM(Hidden Markov Model) speech recognizer. In this method, we use lst-order linear transformation network to fit data of a new speaker to the MLP. Transformation parameters are adjusted by back-propagating classification error to the transformation network while leaving the MLP classifier fixed. The recognition system is based on semicontinuous HMM's which use the MLP as a fuzzy vector quantizer. The experimental results show that rapid speaker adaptation resulting in high recognition performance can be accomplished by this method. Namely, for supervised adaptation, the error rate is signifecantly reduced from 9.2% for the baseline system to 5.6% after speaker adaptation. And for unsupervised adaptation, the error rate is reduced to 5.1%, without any information from new speakers.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 1995.04b
• /
• pp.429-433
• /
• 1995

Compound solid is maded for specially fixed object. A number of compound soild are devided as a unit germetric solid. The special case of rotation about an arbitrary axis makes according selection modelfor transformation. View plane and View region are estiblished for back face removal. After back-face removal eachedge is checked for point of intersection. The designing of of fset surface id maded and tool-path gernerted on the part surface. In tersection point is checked among the offset surfaces. The paper used an efficient algorithm for generating tool paths to apply a notion view volume. Through machining experiments with a 3-axis machining centre, the adequacy of the analysis was confirmed.

• Journal of information and communication convergence engineering
• /
• v.2 no.3
• /
• pp.177-181
• /
• 2004

In this paper, we suggest an Iris recognition system with an excellent recognition rate and confidence as an alternative biometric recognition technique that solves the limit in an existing individual discrimination. For its implementation, we extracted coefficients feature values with the wavelet transformation mainly used in the signal processing, and we used neural network to see a recognition rate. However, Scale Conjugate Gradient of nonlinear optimum method mainly used in neural network is not suitable to solve the optimum problem for its slow velocity of convergence. So we intended to enhance the recognition rate by using Levenberg-Marquardt Back-propagation which supplements existing Scale Conjugate Gradient for an implementation of the iris recognition system. We improved convergence velocity, efficiency, and stability by changing properly the size according to both convergence rate of solution and variation rate of variable vector with the implementation of an applied algorithm.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• v.9 no.1
• /
• pp.816-819
• /
• 2005

This paper proposes a coordinate correction technique using the transformation of 3D parabolic coordinate function and BP(Back Propagation) neural network in order to solve space distortion problem caused by using catadioptric camera. Although Catadioptric camera can obtain omni-directional image at all directions of 360 degrees, it makes an image distorted because of an external form of lens itself. Accordingly, To obtain transformed ideal distance coordinate information from distorted image on 3 dimensional space, we use coordinate transformation function that uses coordinates of a focus at mirror in the shape of parabolic plane and another one which projected into the shape of parabolic from input image. An error of this course is modified by BP neural network algorithm.

• Korean Journal of Applied Biomechanics
• /
• v.13 no.3
• /
• pp.199-215
• /
• 2003

The purpose of this study was to attempt new starting motion and supply present starting motion in the 500M short track speed skating according to the center of gravity position. The center of gravity position ratio was divided starting motion into five(type A : front 80%-back 20%, type B front 70%-back 30%, type C : front 50%-back 50%, type D : front 30%-back 70%, type E : front 20%-back 80%). The three dimension motion analysis with DLT(direct linear transformation) method was executed using two video cameras. The following conclusion was that It was appear that reaction and execution time in starting motion was the most short in type B. It was characteristic that step of skaters was shorten and center of gravity position ratio was not effect to change of the step in each event. It was appear that the displacement of type D and type E were longer than that type A and type B during the starting motion. It was appear that skill types of center of gravity position ratio to the front were lower than that to the back and contract a posture. Observing the above, it was conclusion that skill type B of center of gravity position ratio to the tent was more effect than that to the back. But it is important that these skill type was most used to the competition and estimate the result.