• Proceedings of the KIEE Conference
• /
• 2001.07d
• /
• pp.2663-2665
• /
• 2001

역전파 신경망은 반도체 공정 모델링에 효과적으로 응용이 되고 있으며, 최근 선형뉴런을 비선형 함수 대신 출력층에 이용하여 모델의 예측정확도를 향상 시킨 바 있다. 본 연구에서는 그 원인을 규명하기 위한 모델의 평가지표로서의 유클리디언 웨이트 거리(Euclidean Weight Distance)를 제안한다. 이 지표를 이용하여 신경망의 입력층과 은닉층, 그리고 은닉층과 출력층의 웨이트를 감시하였으며, 그 결과 예측정확도의 향상이 이 지표의 감소에 기인하고 있음을 알았다. 모델링에 이용한 실험데이터는 다중 유도결합형 플라즈마 장비로부터 Langmuir Probe 진단 시스템을 이용하여 수집하였다.

• Journal of IKEEE
• /
• v.22 no.4
• /
• pp.907-915
• /
• 2018

This paper proposes a method to improve the performance of low cost infrared distance sensor. Infrared distance sensor measures the intensity of reflected light and converts it into distance. The proposed method improves the sensing distance of the sensor and makes it operate robustly in various lighting environments. This is achieved by extracting the characteristic curves of the sensor and applying the HDR (High Dynamic Range) technique. The output value of the sensor was obtained by varying the intensity of the infrared input and the exposure time, and the characteristic curve of the sensor was extracted from it.

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.155-161
• /
• 2023

최근 대규모 언어 모델의 활용 방법에 대한 많은 관심이 지속되고 있다. LLM이 생성한 정보의 가장 중요한 도전 과제는 출력 문장이 사실인지 판단하기 어렵다는 점이다. 본 논문은 하나 또는 복수의 LLM를 구조화하여, 생산되는 다양한 결과를 활용하는 방안을 탐구한다. LLM이 생성한 출력 정보를 고려하여 사실 관계 확인 과제를 수행함에 있어서, 성능 향상 가능성을 평가하기 위한 실험을 진행했다. 대규모 언어모델의 구조화를 통해 입력 정보의 제어를 할 경우, 기존 최고 성능보다 4.75의 정확도 향상을 관찰할 수 있었다.

• Journal of Internet Computing and Services
• /
• v.6 no.1
• /
• pp.105-114
• /
• 2005

This paper describes a new measuring technique by analysing multiple time-series patterns. This paper's goal is that extracts a really measured value having a sample pattern which is the best matched with an inputted time-series, and calculates a difference ratio with the value. Therefore, the proposed technique is not a recognition but a measurement. and not a hardware but a software. The proposed technique is consisted of three stages, initialization, learning and measurement. In the initialization stage, it decides weights of all parameters using importance given by an operator. In the learning stage, it classifies sample patterns using LBG and DTW algorithm, and then creates code sequences for all the patterns. In the measurement stage, it creates a code sequence for an inputted time-series pattern, finds samples having the same code sequence by hashing, and then selects the best matched sample. Finally it outputs the really measured value with the sample and the difference ratio. For the purpose of performance evaluation, we tested on multiple time-series patterns obtained from etching machine which is a semiconductor manufacturing.

• The KIPS Transactions:PartC
• /
• v.10C no.4
• /
• pp.463-470
• /
• 2003

Many types of switching fabrics have been proposed for use in ATM networks. Multistage Interconnection Networks (MINs) constitute a large class of ATM switching systems that are widely used in today´s internetworking. One of the most veil-known types of multistage networks is the banyan network. The banyan network is attractive for its simple routing scheme and low hardware complexity, but its throughput is very limited due to internal blocking and output contention. In this paper, we propose an input-buffered dual-banyan switch model with multiple switching fabric between switch input and output to avoid internal and Head-of Line blocking. By performance analysis and simulation, we show that our model has a lower ceil delay and 96% throughput which is much better than other banyan-type switch architecture.

• Journal of Korea Multimedia Society
• /
• v.12 no.2
• /
• pp.209-218
• /
• 2009

3D backprojection is a kind of reconstruction algorithm to generate volume data consisting of tomographic images, which provides spatial information of the original 3D data from hundreds of 2D projections. The computational time of backprojection increases in proportion to the size of volume data and the number of projection images since the value of every voxel in volume data is calculated by considering corresponding pixels from hundreds of projections. For the reduction of computational time, fast GPU based 3D backprojection methods have been studied recently and the performance of them has been improved significantly. This paper presents two multiple GPU based methods to maximize the parallelism of GPU and compares the efficiencies of two methods by considering both the number of projections and the size of volume data. The first method is to generate partial volume data independently for all projections after allocating a half size of volume data on each GPU. The second method is to acquire the entire volume data by merging the incomplete volume data of each GPU on CPU. The in-complete volume data is generated using the half size of projections after allocating the full size of volume data on each GPU. In experimental results, the first method performed better than the second method when the entire volume data can be allocated on GPU. Otherwise, the second method was efficient than the first one.

• Journal of Navigation and Port Research
• /
• v.44 no.3
• /
• pp.145-150
• /
• 2020

The ship motion of large LNG ships affects ships' safety. The purpose of this study was to estimate the transfer function of roll motion among the hull motion of 153,000 m3 class LNG vessels. The ship motion transfer function was modeled using a Linear Time-Invarient system with single input, single output, and transfer function. The transfer function of the ship motion was estimated by the system identification method using single ocean wave as input of the model, and using the roll motion of the LNG ship obtained through ANSYS as the output of the model. The usefulness of the experimental results was evaluated using the precision and estimation rate of the model for cases wherein the different transfer function dimensions. Results of the experiment showed a precision at 99% and 98%, with estimation rate at 78% and 50%. From these results, we found the proposed method of estimating the transfer function of ship motion in this study reasonable. In the future, data of ship motion in actual sea conditions will be acquired and it will be applied to make the construction of models with multiple inputs and multiple outputs for practical use.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.35 no.11
• /
• pp.990-998
• /
• 2007

Adaptive output feedback control technique using Neural Networks(NN) is proposed for uncertain nonlinear Multi-Input Multi-Output(MIMO) systems. Modified Dynamic Inversion Model(MDIM) is introduced to decouple uncertain nonlinearities from inversion-based control input. MDIM consists of approximated dynamic inversion model and inversion model error. One NN is applied to compensate the MDIM of the system. The output of the NN augments the tracking controller which is based upon a filtered error approximation with online weight adaptation laws which are derived from Lyapunov's direct method to guarantee tracking performance and ultimate boundedness. Several numerical results are illustrated in the simulation of Van der Pol system and unmanned helicopter with model uncertainties.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.30 no.4C
• /
• pp.207-214
• /
• 2005

In the indoor wireless communication, it is considered the indoor wireless system architecture of single input multiple output (SIMO) that used with multiple antenna in order to cope with the indoor fading characteristic due to severe angler spread. We propose the mean steering vector technique as a method to enhance the system Performance, implement the test bed system composed of a PC and the algorithms of the wireless system, and analysis the performance of those algorithms. In addition, the overall operation scenario, overall architecture, and the execution time of the algorithms, of the test bed for the indoor wireless system are presented.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.23 no.2
• /
• pp.229-234
• /
• 2012

This paper presents channel measurements for wireless body area network(WBAN) and provides performance evaluation from the measurement. We measured the radio propagation in 2.45 GHz ISM band in an anechoic chamber according to various human movements and the position of transmit antennas. Two transmit antennas are mounted on different positions of human body for the purpose of comparing the diversity gain and correlation between the channels in $2{\times}1$ multiple-input single-output(MISO) systems. The experimental results show that the outage capacity is closely related with the correlation coefficient between channels in transmit diversity system.