• 정보학연구
• /
• 제8권1호
• /
• pp.51-56
• /
• 2005

The Back propagation algorithm is a very effective supervised training method for multi-layer feed forward neural networks. This paper studies the decision boundary formation based on the Back propagation algorithm. The discriminating powers of several neural network topology are also investigated against five manually created data sets. It is found that neural networks with multiple hidden layer perform better than single hidden layer.

• Restorative Dentistry and Endodontics
• /
• 제16권1호
• /
• pp.167-180
• /
• 1991

The purpose of theis study was to evaluate the effectiveness of ultrasonic instrumentation in removing debris and smear layer from the root canal walls. 54 extracted, permanent single rooted teeth were randomly divided into 6 groups of 9 teeth. After canal preparation, the teeth were longitudinally sectioned. The sepcimens were then naturally dried for 2 days, given a maximun thickness gold coating, and examined under the SEM (JSM-35C type, JAPAN). Photographs of all specimens were then taken of the middle and the apical third of the root canal wall. The results were as follows: 1. In all groups, debris and the smear layer were not completely removed from the canal wall 2. There were no significant differences between at the apical third and at the middle third in removing debris and the smear layer in all groups. 3. There were no significant difference between the step - back group and the ultrasound group in removing debris and smear layer. 4. In general, the step - back/ultrasonund groups showed greater canal debridement than the step - back group or ultrasound group. 5. The step - back/ultrasound group with a No. 25 file for 3 min. showed significantly greater canal debridement than the step-back group (p<0.05), or the other step - back/ultrasound groups(p<0.05).

• 한국지반공학회논문집
• /
• 제25권9호
• /
• pp.93-100
• /
• 2009

본 연구에서 역해석에 이용한 직접법은 실제 변위와 목적함수의 차이를 최소화하고 미지변수를 보정하는 최적화가 핵심 과정이다. 연구과정이 복수의 지층을 단계별로 역해석하여 각 지반의 미지변수를 역해석하기 때문에 지층 간 미지변수의 간섭 효과를 최소화 할 수 있는 추가적인 최적화가 필요하다. 따라서 효율적이고 정확한 최적화를 위해서 목적함수의 편미분 계산과정이 없이 최적해를 구할 수 있는 직접탐색법(direct search method)을 사용하였다. 본 연구의 목적은 지하구조체가 포함된 지층을 상부와 하부의 지층으로 모형화하고, 다중 미지변수를 역해석하는 기법을 개발하는 데에 목적을 둔다. 또한 단계별 역해석된 미지변수의 상호 간섭 오차를 최소화하기 위해서 반복 역해석을 실시하고, 실제 존재하는 터널에 대하여 적용성에 대해 검토하였다. 그 결과, 단층 모형화보다 다층 모형화 결과가 실제 지하구조물거동을 정확히 나타냈으며 다층 경계부에 위치한 지하구조물의 다층지반 모형화를 통한 역해석이 유효함을 확인하였다.

• Journal of the Korean Wood Science and Technology
• /
• 제27권3호
• /
• pp.82-90
• /
• 1999

송광사 삼세불화는 조성년도가 확실하지 않지만 남아있는 서기(書記)와 재질 분석을 통해 19세기초에 제작된 것으로 추정 할 수 있었다. 그리고 공시불화의 보존처리를 위해 안료층과 바탕층, 지지층을 구성하고 있는 재질을 분석하고 이에 적합한 지지층인 배첩지를 초지하여 그 적합성을 시험한 바 다음과 같은 결론을 얻었다. 1. 삼세불화의 단면구조는 안료층과 명주로 된 바탕층 그리고 한지(韓紙)로 이루어진 지지층으로 구성되어 있어 19세기 당시의 불화는 기본적으로 안료층, 바탕층, 지지층으로 구성되어 있음을 알 수 있었다. 또한 안료층과 바탕층의 접착은 아교를 사용하였고 지지층의 sheet 배첩은 호료(糊料)를 사용했음을 알 수 있었다. 2. 삼세불화의 지지층은 5겹으로 구성되어 있고 두 종류의 종이가 사용되었는데 바탕층과 접해 있는 1~2층은 순수한 닥섬유로 제조한 한지(韓紙)를 사용하였고 나머지 3개 층은 폐마섬유(廢麻纖維)에 닥섬유가 혼합된 종이를 사용하였다. 더 많은 자료가 확보되어야만 단언할 수 있지만 재질을 통해 당시의 상황을 파악해 보면 19세기 당시의 사회는 경제적으로 어려움을 겪고 있었고 더욱이 사찰 제지업은 피폐해져 운영이 어려워 후면에 사용한 배첩지는 질이 떨어지는 폐마를 이용 했을 것으로 추측되었다. 3. 본 재료의 바탕층과 지지층간의 인장력을 고려하여 새롭게 제조해 보수용으로 사용한 바 본 공시재료의 보수용 배첩지로서 적합했음을 알 수 있었다.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2000년도 추계종합학술대회 논문집(1)
• /
• pp.441-444
• /
• 2000

This paper covers the performance testing for layer 3 Ethernet switch based on various methodologies by which we can measure essential metrics such as throughput, latency, frame loss rate, and back to back frames. In the first place, layer 2 and layer 3 switch evolution is introduced followed by description of IP packet switching in layer 3 switch. And then, the above test metrics and test methodologies are illustrated as well. At last, we conduct the performance testing for layer 3 switch in case of transmitting packets of 64, 128, 256, 512, 1024, 1280, and 1518 byte size and analyze then results.

• 전자공학회논문지B
• /
• 제29B권4호
• /
• pp.58-65
• /
• 1992

In this paper new back propagation algorithm which adds neurons to hidden layer is proposed. this proposed algorithm is applied to the pattern recognition of written number coupled with back propagation algorithm through omitting redundant learning. Learning rate and recognition rate of the proposed algorithm are compared with those of the conventional back propagation algorithm and the back propagation through omitting redundant learning. The learning rate of proposed algorithm is 4 times as fast as the conventional back propagation algorithm and 2 times as fast as the back propagation through omitting redundant learning. The recognition rate is 96.2% in case of the conventional back propagation algorithm, 96.5% in case of the back propagation through omitting redundant learning and 97.4% in the proposed algorithm.

• Journal of Magnetics
• /
• 제12권3호
• /
• pp.118-123
• /
• 2007

We have studied the effect of Au-Cu back layer system ${\sim}10{\AA}$ thick on the properties of a spin valve. The back layers were Cu, Au, co-sputtered $Cu_xAu_{1-x}$, laminated $[Au/Cu]_n$. and bi-layer [Au/Cu]. When Au was added to the Cu, the resistance of the spin valve abruptly increased most likely due to impurity scattering. The GMR values were not increased significantly for all the structures. In the case of co-sputtered $Cu_xAu_{1-x}$, the changes in the resistance, ${\Delta}R$, was increased at a composition of ${\sim}Au_{0.5}Cu_{0.5}$. This increase in ${\Delta}R$ is due to increase in the resistance and not from the enhanced spin-dependent scattering. The structural analyses showed that the orthorhombic $Au_{0.5}Cu_{0.5}$ was formed in the back layer instead of the face-centered tetragonal $Au_{0.5}Cu_{0.5}$ as we expected. Thermal annealing over $400^{\circ}C$ may be required to have face-centered tetragonal in the $10{\AA}$ thick ultra-thin film. In the case of a laminated or bi-layered back layer, the properties of the spin valve were improved, which may be attributed to the increase in the mean free path of conduction electrons.

• 한국자기학회지
• /
• 제4권1호
• /
• pp.20-24
• /
• 1994

CoCr/NiFe 이중매체의 기록, 재생특성은 CoCr 기록층뿐만 아니라 NiFe연자성층의 자기적 특성에도 강하게 영향을 받는다. NiFe/Ge의 다층화로 CoCr/NiFe 매체의 기록감도와 재생전압을 증 가시킬 수 있었으며 이것은 연자성층의 높은 투자율과 CoCr기록층의 높은 수직이방성에 기인하였다. 그러나 연자성층의 투자율의 향상이 기록감도의 현저한 향상에 비해 재생전압의 현저한 증가를 가 져오지 못하는데 이것은 유한요소수치해석에서도 확인할 수 있었다. 반면에 NiFe/Ge 다층연자성박 막을 사용한 CoCr/NiFe 매체의 피크이동(peak shift)특성은 NiFe 단층연자성박막을 사용한 CoCr/NiFe 매체에 비해 악화되었다. 그러나 CoCr층과 NiFe층 사이에 10 nm의 CoZrNb 박막을 삽입시키면 높은 재생전압과 낮은 파크이동을 동시에 얻을 수 있었다. 이 연구에서 는 이러한 기록, 재생특성을 미 세구조와 연관시켜 검토하였다.

• 한국표면공학회지
• /
• 제15권2호
• /
• pp.69-76
• /
• 1982

With a view to manufacturing membranes for separation of gas mixtures, Al foils were anodized in a 2% oxalic-acid electrolyte at 40V and 80V. When anodizing was completed and Barrier layer existed at the extreme back site of the foil, the anodized foil was made to react with only electrolyte, with switching off the electric power. When the size and density of pores were changed through voltage change, the membr-anes did not show large difference in the permeability. Reacting with electrolyte, the existing Barrier layer turns into porous layer. During this process, several small pores grow from one relatively large pore, getting to the back site. The number and size of the small pores getting to the back surface increase as time passing. This change of Barrier layer into porous layer is thought to be directly related to the permeability change of the membranes. The selectivity of an anodized Al membrane was not related to the voltage change, and was high, being similar to the theoretical selctivity of metallic membranes, according to my observation.

• 한국도로학회논문집
• /
• 제17권5호
• /
• pp.25-31
• /
• 2015

PURPOSES: The objective of this study is to analyze the relationship between the FWD back-calculated modulus and dynamic modulus of asphalt layers for existing asphalt pavements. METHODS: To evaluate the dynamic modulus of the asphalt mixture in the existing and new asphalt layers, the uniaxial direct tension test was conducted on small asphalt specimens obtained from the existing asphalt-covered pavements. A dynamic modulus master curve was estimated by using the uniaxial direct tension test for each asphalt layer. The falling weight deflectometer (FWD) testing was conducted on the test sections, and the modulus values of pavement layers were back-calculated using the genetic algorithm and the finite element method based back-calculation program. The relationship between measured and back-calculated asphalt layer moduli was examined in this study. The normalized dynamic modulus was adopted to predict the stiffness characteristics of asphalt layers more accurately. RESULTS: From this study, we can conclude that there is no close relationship between dynamic modulus of first layer and back-calculated asphalt modulus. The dynamic moduli of second and third asphalt layers have some relation with asphalt stiffness. Test results also showed that the normalized dynamic modulus of the asphalt mixture is closely related to the FWD back-calculated modulus with 0.73 of R square value. CONCLUSIONS: The back-calculated modulus of asphalt layer can be used as an indicator of the stiffness characteristics of asphalt layers in the asphalt-covered pavements.