• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.39 no.4
• /
• pp.19-27
• /
• 2002

We proposed an auto-tuning method of learning rate for performance improvement of backpropagation algorithm. Proposed method is used a fuzzy logic system for automatic tuning of learning rate. Instead of choosing a fixed learning rate, the fuzzy logic system is used to dynamically adjust learning rate. The inputs of fuzzy logic system are ${\Delta}$ and $\bar{{\Delta}}$, and the output is the learning rate. In order to verify the effectiveness of the proposed method, we performed simulations on a N-parity problem, function approximation, and Arabic numerals classification. The results show that the proposed method has considerably improved the performance compared to the backpropagation, the backpropagation with momentum, and the Jacobs' delta-bar-delta.

• Journal of the Korean Geotechnical Society
• /
• v.19 no.1
• /
• pp.151-161
• /
• 2003

In this study, an interpretation method is proposed to evaluate the stress conditions, including individual excess pore pressure components ($\Delta{u}_{oct}$/ and $\Delta{u}_{shear}$), of normally consolidated clay elements adjacent to the cone face. It is assumed that the stress path of triaxial compression is representative f3r that of the soil element and the soil is elastic-perfectly plastic material. The proposed method is applied to the results of miniature piezocone tests conducted at Louisiana State University calibration chamber system. Based on the results of interpretation, it was found that the ratio of $\Delta{u}_{oct}$/ $\Delta{u}$and $\Delta{u}_{shear}$/$\Delta{u}$ estimated by the proposed method is affected only by the pore pressure parameter. The proposed method gives consistent and reliable values of $\Delta{u}_{oct}$/ $\Delta{u}$and $\Delta{u}_{shear}$/$\Delta{u}$ compared with early works, whereas those obtained by other solutions are significantly dependent on the accuracy in estimating soil properties such as undrained shear strength and rigidity index.

• Journal of The Korean Astronomical Society
• /
• v.33 no.2
• /
• pp.89-95
• /
• 2000

As an efficient method to detect blending of general gravitational microlensing events, it is proposed to measure the shift of source star image centroid caused by microlensing. The conventional method to detect blending by this method is measuring the difference between the positions of the source star image point spread function measured on the images taken before and during the event (the PSF centroid shift, ${\delta}{\theta}$c,PSF). In this paper, we investigate the difference between the centroid positions measured on the reference and the subtracted images obtained by using the difference image analysis method (DIA centroid shift, ${\delta}{\theta}$c.DIA), and evaluate its relative usefulness in detecting blending over the conventional method based on ${\delta}{\theta}$c,PSF measurements. From this investigation, we find that the DIA centroid shift of an event is always larger than the PSF centroid shift. We also find that while ${\delta}{\theta}$c,PSF becomes smaller as the event amplification decreases, ${\delta}{\theta}$c.DIA remains constant regardless of the amplification. In addition, while ${\delta}{\theta}$c,DIA linearly increases with the increasing value of the blended light fraction, ${\delta}{\theta}$c,PSF peaks at a certain value of the blended light fraction and then eventually decreases as the fraction further increases. Therefore, measurements of ${\delta}{\theta}$c,DIA instead of ${\delta}{\theta}$c,PSF will be an even more efficient method to detect the blending effect of especially of highly blended events, for which the uncertainties in the determined time scales are high, as well as of low amplification events, for which the current method is highly inefficient.

• 제어로봇시스템학회:학술대회논문집
• /
• 1989.10a
• /
• pp.82-86
• /
• 1989

In this paper, digital autopilot design methods are investigated and a new method is suggested in order to improve existing problems. The method is based on .delta. transform (1) and overcome numerical problems occurring in the process of discretization. We illustrate design procedures using .delta. transform and suggest a hardware and software structure for digital autopilot implemented by microprocessor.

• Journal of Korean Society of Occupational and Environmental Hygiene
• /
• v.4 no.1
• /
• pp.17-24
• /
• 1994

The urinary excretion of ${\delta}$-aminolevulinic acid has been widely used as a measure of the biological effect of lead in lead exposed workers. It is usually measured by colorimetric method based on the color reaction of ALA-pyrrole with Ehrlich's reagent. But the results of ${\delta}$-ALA in urine by this method are somewhat artificially higher than expected due to the urinary ALA-like compound such as aminoacetone. On the other hand, the recently developed fluorometric HPLC method is very sensitive and specific for the measuring urinary ALA. In order to compare the data obtained by two methods and to investigate the interrelation between two methods, 117 lead workers with different lead exposure were checked urinary ${\delta}$-ALA, blood lead and other lead exposure related indices. The results obtained are as follows; 1. Urinary excretion of ${\delta}$-ALA by colorimetric method is 2.15 times higher than HPLC method in overall, revealing 2.47 times in workers of blood lead less than $20{\mu}g/dl$, 2.53 times in workers of blood lead $21-40{\mu}g/dl$ and 1.86 times in workers of blood lead over $41{\mu}g/dl$, respectively. 2. While the correlation coefficients of ${\delta}$-ALA measured by colorimetric method with blood lead and blood ZPP was 0.571 and 0.629, those of ${\delta}$-ALA measured by HPLC with blood lead and blood ZPP were 0.6l0 and 0.637. All the correlation coefficients were statistically significant, but there was no statistical difference of correlation coefficients between two methods. 3. The correlation coefficient of urinary excretion of ${\delta}$-ALA between two method was 0.838 without any correction, but it was 0.852 with the correction of specific gravity 1.024. 4. Simple linear regression of ${\delta}$-ALA measured by HPLC method on ${\delta}$-ALA measured by colorimetric method was (ALA-UPH)=-0.245+0.536 (ALA-UCO) without any correction and it was (SP ALA)=-0.525+0.598 (SP ALA-UCO) with the correction of specific gravity 1.024. With above results, it is recommended that the diagnostic criteria of ${\delta}$-ALA for lead poisoning needed to be revised if ${\delta}$-ALA is measured by HPLC rather than colorimetric method.

• Proceedings of the IEEK Conference
• /
• 2003.11a
• /
• pp.19-22
• /
• 2003

This paper proposes the method of visual resolution enhancement to render a color image on a delta-structured display. The proposed method adopted a subpixel rendering method to reduce a color fringe error caused by delta- structured display and to improve visual resolution

• Journal of the Korean Ceramic Society
• /
• v.33 no.3
• /
• pp.327-331
• /
• 1996

$\delta$-FeOOH was prepared by rapid oxidation method of Fe(OH)2 using H2O2. The effects of reaction temperature and mole ratio ([OH-]/[Fe2+])의 몰비를 제조 변수로 하여 최종 생성된 $\delta$-FeOOH의 입자크기 입자형태, 자기특성을 조사하였다. Fe(OH)2 의 반응온도 및 [OH-]/[Fe2+] 비가 $\delta$-FeOOH의 입자크기 및 형상에 많은 영향을 미침을 알수 있었으며 입자 크기는 이 두인자에 비례하여 증가하였다 Fe(OH)2 의 반응온도가 4$0^{\circ}C$ [OH-]/[Fe2+]=5 Fe(OH)2 숙성 시간 2시간에서 제조된 $\delta$-FeOOH를 TEM, VSM으로 입자의 크기 및 자기특성을 조사한결과 평균 입경이 630$\AA$ 정도이고 입도 분포가 양호하였으며 포화자화 및 보자력은 각각 20.8emu/g 210 Oe였다.

• Journal of information and communication convergence engineering
• /
• v.1 no.3
• /
• pp.157-162
• /
• 2003

We propose a learning method for improving the performance of the backpropagation algorithm. The proposed method is using a fuzzy logic system for automatic tuning of the learning rate of each weight. Instead of choosing a fixed learning rate, the fuzzy logic system is used to dynamically adjust the learning rate. The inputs of fuzzy logic system are delta and delta bar, and the output of fuzzy logic system is the learning rate. In order to verify the effectiveness of the proposed method, we performed simulations on the XOR problem, character classification, and function approximation. The results show that the proposed method considerably improves the performance compared to the general backpropagation, the backpropagation with momentum, and the Jacobs'delta-bar-delta algorithm.

• The Transactions of the Korean Institute of Electrical Engineers A
• /
• v.48 no.10
• /
• pp.1231-1238
• /
• 1999

This paper presents an approach for the optimal heating load Identification using Diagonal Recurrent Neural Networks(DRNN). In this paper, the DRNN captures the dynamic nature of a system and since it is not fully connected, training is much faster than a fully connected recurrent neural network. The architecture of DRNN is a modified model of the fully connected recurrent neural network with one hidden layer. The hidden layer is comprised of self-recurrent neurons, each feeding its output only into itself. In this study, A dynamic backpropagation (DBP) with delta-bar-delta learning method is used to train an optimal heating load identifier. Delta-bar-delta learning method is an empirical method to adapt the learning rate gradually during the training period in order to improve accuracy in a short time. The simulation results based on experimental data show that the proposed model is superior to the other methods in most cases, in regard of not only learning speed but also identification accuracy.

• Journal of the Korean Ceramic Society
• /
• v.35 no.10
• /
• pp.1094-1100
• /
• 1998

The thermal fatigue behavior of alumina ceramics was investigated by water quenching method. Single-quench thermal shock tests were performed to decide the critical thermal shock temperature difference ($\Delta$Tc) which was found to be 225$^{\circ}C$ Cyclic thermal shock fatigue tests were performed at temperature diff-erences of 175$^{\circ}C$, 187$^{\circ}C$ and 200$^{\circ}C$ respectively. After cyclic thermal shock fatigue test the distributions of retained strength and crack were observed. Retained strength was measured by four point bending method and crack observation method bydye penetration. In terms of the retained strength distribution the critical number of thermal shock cycles(Nc) were 7 for $\Delta$T=200$^{\circ}C$, 35 for $\Delta$T=187$^{\circ}C$ and 180for $\Delta$T=175$^{\circ}C$ respec-tively. In terms of the crack observation the critical number of thermal shock cycles were 5 for $\Delta$T==200$^{\circ}C$ 20 for $\Delta$T==187$^{\circ}C$ and 150 for $\Delta$T=175$^{\circ}C$ respectively. The difference of Nc investigated by two different methods is due to the formation of the longitudinal cracks which had no effect on the four point bending strength. Therefore the thermal fatigue behavior of alumina ceramics could be more accurately described by the crack observation method than the retained strength measurement method.