• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.60 no.2
• /
• pp.352-356
• /
• 2011

Transparent conducting oxide films like ITO/Au/ITO and AZO/Au/AZO were fabricated with a sputter at a low-temperature of less then $70^{\circ}C$ and their crystallization and opto-electrical properties were studied. X-ray diffractiometry showed that single-ITO layer was amorphous, whereas, ITO of ITO/Au/ITO multi-layer was crystal. The ITO crystallization and its orientation depended on Au crystallization. Surface roughness of the ITO-multi-layers were in the range of 29-88% of that of ITO-single layer. ITO on amorphous gold layer had more rough surface than ITO on crystal gold. The gold layer between ITO improved electrical conductivity. Carrier density, mobility, resistivity and sheet resistance of ITO-single layer were $2.3{\times}10^{19}/cm^3$, $85{\times}cm^2$/Vs, $31{\times}10^{-4}{\Omega}cm$, and $310{\times}{\Omega}/cm^2$, respectively. Those of ITO/Au/ITO-multi-layers depended on Au-interlayer-thickness, which were in the range of $3.6{\times}10^{19}{\sim}4.2{\times}10^{21}/cm^3$, $43{\sim}85cm^2$/Vs, $0.17{\times}10^{-4}{\sim}25{\times}10^{-4}{\Omega}cm$, and $1.7{\sim}20{\times}{\Omega}/cm^2$, respectively. The sheet resistances of the single-layer ITO and the multi-layer ITO were 310 and $2.7{\sim}21{\Omega}/cm^2$, respectively. That of AZO/Au/AZO was $8.6{\Omega}/cm^2$, which was better than the single-layer ITO.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.05a
• /
• pp.11-12
• /
• 2023

GPGPU 환경에서의 ML 모델이 다양한 분야에 지속적으로 활용되면서, 이미지 분할(image segmentation) 연구가 활발하다. multi-GPU 환경에서 성능 최적화를 위하여 병렬화 기법들이 활용되고 있다. 본 연구에서는 multi-GPU 환경에서 U-Net 모델의 전체 수행 시간을 단축하기 위해 convolution 연산을 최적화하는 기법을 적용하는 실험을 진행하였고 shared memory, data parallelism 를 적용하여 82% 성능 향상을 보여주었다.

• Journal of Electrical Engineering and Technology
• /
• v.10 no.2
• /
• pp.699-707
• /
• 2015

Although digit character recognition has got a significant improvement in recent years, it is still challenging to achieve satisfied result if the data contains an amount of distracting factors. This paper proposes a novel digit character recognition approach using a multi-layer hierarchical model, Hybrid Restricted Boltzmann Machines (HRBMs), which allows the learning architecture to be robust to background distracting factors. The insight behind the proposed model is that useful high-level features appear more frequently than distracting factors during learning, thus the high-level features can be decompose into hybrid hierarchical structures by using only small label information. In order to extract robust and compact features, a stochastic 0-1 layer is employed, which enables the model's hidden nodes to independently capture the useful character features during training. Experiments on the variations of Mixed National Institute of Standards and Technology (MNIST) dataset show that improvements of the multi-layer hierarchical model can be achieved by the proposed method. Finally, the paper shows the proposed technique which is used in a real-world application, where it is able to identify digit characters under various complex background images.

• Proceedings of the Korean Vacuum Society Conference
• /
• 2013.02a
• /
• pp.618-618
• /
• 2013

Graphene is only one atom thick planar sheet of sp2-bonded carbon atoms arranged in a honeycomb crystal lattice, which has flexible and transparent characteristics with extremely high mobility. These noteworthy properties of graphene have given various applicable opportunities as electrode and/or channel for various flexible devices via suitable physical and chemical modifications. In this work, for the development of all-graphene devices, we performed to synthesize alternately patterned structure of mono- and multi-layer graphene by using the patterned Ni film on Cu foil, having much different carbon solid solubilities. Depending on the process temperature, Ni film thickness, introducing occasion of methane and gas ratio of CH4/H2, the thickness and width of the multi-layer graphene were considerably changed, while the formation of monolayer graphene on just Cu foil was not seriously influenced. Based on the alternately patterned structure of mono- and multi-layer graphene as a channel and electrode, respectively, the flexible TFT (thin film transistor) on SiO2/Si substrate was fabricated by simple transfer and O2 plasma etching process, and the I-V characteristics were measured. As comparing the change of resistance for bending radius and the stability for a various number of repeated bending, we could confirm that multi-layer graphene electrode is better than Au/Ti electrode for flexible applications.

• Proceedings of the Korean Institute of Electrical and Electronic Material Engineers Conference
• /
• 2002.07b
• /
• pp.653-656
• /
• 2002

In this paper, we present the simulation results of multi-layer VCO(voltage controlled oscillator), which is composed of resonator, oscillator, and buffer circuit, using EM simulator and nonlinear RF circuit simulator. EM simulator is used for obtaining the EM(Electromagnetic) characteristics of conductor pattern as well as designing the multi-layer VCO. Obtained EM characteristics were used as real components in nonlinear RF circuit simulation. Finally the overall VCO was simulated by the nonlinear RF circuit simulator. The material for the circuit pattern was Ag and the dielectric was DuPont 951AT, which will be applied for LTCC process. The structure of multi-layer VCO is constructed with 4 conducting layer. Simulated results showed that the output level was about 4.5 [dBm], the phase noise was -104 [dBc/Hz] at 30 [kHz] offset frequency, the harmonics -8 dBc, and the control voltage sensitivity of 30 [MHz/V] with a DC current consumption of 9.5 [mA]. The size of VCO is $6{\times}9{\times}2$ mm(0.11[cc]).

• Bulletin of the Korean Chemical Society
• /
• v.24 no.2
• /
• pp.183-188
• /
• 2003

The micro-patterning of multi-layered thin films containing CdS and $Ta_2O_5$ layers on ITO substrate with various structures was successfully obtained by combining three different techniques: chemical solution depositions, sol-gel, and microcontact printing (μCP) methods using octadecyltrichlorosilane (OTS) as the organic thin layer template. $Ta_2O_5$ layer was prepared by sol-gel casting and CdS one obtained by chemical solution deposition, respectively. Parallel and cross patterns of multi-layers with $Ta_2O_5$ and CdS films were fabricated additively by successive removal of OTS layer pre-formed. This study presents the designed architectures consisting of the two types of feature having horizontal dimensions of 170 ㎛ and 340 ㎛ with constant thickness ca. 150 nm of each deposited materials. The thin film lay-out of the cross-patterning is composed of four regions with chemically different layer compositions, which are confirmed by Auger electron microanalysis.

• Journal of the Korean Institute of Telematics and Electronics A
• /
• v.29A no.9
• /
• pp.46-55
• /
• 1992

This paper propose how to calculate the capacitance for VLSI interconnection lines in multi-dielectric layer. The proposed method is a expansive form of 3-dimensional direct intergral method developed in single-dielectric layer. We took into consideration the effect of multi-dielectric layer by using additional boundary condition instead of modified Green's function. It is used the potential equations in line surface and the electric field equations in dielectric interface as the boundary condition. RC delay time for interconnection line of multi-dielectric layer is obtained from the calculated capacitance value. At this time, we are used Al and WSiS12T as interconnection materials.

• Journal of the Korean institute of surface engineering
• /
• v.39 no.1
• /
• pp.9-12
• /
• 2006

Al-doped ZnO(AZO)/Ag/AZO multi-layer films deposited on PET substrate by RF magnetron sputtering have a much better electrical properties than Al-doped ZnO single-layer films. The multi-layer structure consisted of three layers, AZO/Ag/AZO, the optimum thickness of Ag layers was determined to be $112{\AA}$ for high optical transmittance and good electrical conductivity. With about $1800{\AA}$ thick AZO films, the multi-layer showed a high optical transmittance in the visible range of the spectrum. The electrical and optical properties of AZO/Ag/AZO were changed mainly by thickness of Ag layers. A high quality transparent electrode, having a resistance as low as $6\;W/{\square}$ and a high optical transmittance of 87% at 550 nm, was obtained by controlling Ag deposition parameters.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.35 no.5A
• /
• pp.474-481
• /
• 2010

In this paper, we present an effective cross layer-based video transmission algorithm over multi-hop mobile ad hoc networks. The proposed algorithm selects the most efficient PHY mode of wireless LAN multi-rate service at each node in a distributed way based on the available information at application, MAC, and physical layers in order to minimize end-to-end delay and maintain packet loss rate in tolerable range at the receiver. Finally, experimental results are provided to show superior performance of the proposed algorithm.

• Journal of Korean Institute of Industrial Engineers
• /
• v.28 no.2
• /
• pp.147-161
• /
• 2002

In novelty detection, one attempts to discriminate abnormal patterns from normal ones. Novelty detection is quite difficult since, unlike usual two class classification problems, only normal patterns are available for training. Auto-Associative Multi-Layer Perceptron (AAMLP) has been shown to provide a good performance based upon the property that novel patterns usually have larger auto-associative errors. In this paper, we give a mathematical analysis of 2-layer AAMLP's output characteristics and empirical results of 2-layer and 4-layer AAMLPs. Various activation functions such as linear, saturated linear and sigmoid are compared. The 2-layer AAMLPs cannot identify non-linear boundaries while the 4-layer ones can. When the data distribution is multi-modal, then an ensemble of AAMLPs, each of which is trained with pre-clustered data is required. This paper contributes to understanding of AAMLP networks and leads to practical recommendations regarding its use.