• Electronics and Telecommunications Trends
• /
• v.21 no.6 s.102
• /
• pp.124-132
• /
• 2006

무선 통신 시스템에서 다중 경로 감쇠로 인한 심볼의 크기와 위상의 왜곡이 일어나는데 이를 추정하여 보상하기 위해 채널 추정 기법이 사용된다. OFDM 시스템에서의 채널 추정 기법에는 훈련 심볼이나 파일럿을 사용하여 채널을 추정하는 기법과 파일럿을 사용하지 않는 Blind 채널 추정 기법 등이 있다. 본 고에서는 I장 개요에 이어 II장에서는 훈련 심볼을 이용한 채널 추정 기법을 WPAN과 WLAN을 예로 들어 설명하고 LS,1-D LMMSE, Low rank LMMSE 알고리듬에 대해 설명한다. III장에서는 PSAM 채널추정에 대해 1차원 PSAM 기법과 2차원 파일럿 패턴과 Wiener filtering을 이용한 2-D LMMSE, Low rank LMMSE, Separable Wiener filter에 대하여 설명한다. IV장에서는 ESAE와 Blind 채널 추정 기법을 간략히 소개하고, V장에서 채널 추정의 최신 연구 동향을 소개한다.

• Journal of Fisheries and Marine Sciences Education
• /
• v.24 no.5
• /
• pp.653-661
• /
• 2012

The purpose of this research is to analyze teacher's lesson language pattern based on robot class. For this research as an analytical tool AF (Advanced Flanders) was utilized. Actually ClassReport ver 1.0 computer program was used in the process of input data. From the results of the Flanders index, The major instruction sequences were 4-8-3 in elementary school and 4-8 in middle school. The teacher's remarks in robot class in the elementary school rank 'instruction', 'question'. And In the middle school rank 'instruction' and 'positive advice' are very high ratio, but 'indication' is low ratio. The teacher constantly teach and ask question ratio in the middle school was higher than elementary school. But a tendency for non-indication was low ratio in the middle school than elementary school. These results could provide effective cues and information on how to to improve instruction.

• 한국연소학회:학술대회논문집
• /
• 2012.04a
• /
• pp.205-208
• /
• 2012

In this study, the thermal behavior and combustion characteristics of different ranks of coals and their blends were investigated to obtain information necessary for the evaluation of the co-processing of blends with low-rank coals. Thermogravimetric analysis(TGA) and differential thermal analysis(DTA) were carried out at different temperature from ambient temperature to $800^{\circ}C$, and a laboratory-scale pulverized coal combustion burner was used with coal feeing rate of $1.04{\times}10^{-4}kg/s$.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2019.06a
• /
• pp.84-85
• /
• 2019

MPEG-NNR (Compression of Neural Network for Multimedia Content Description and Analysis) aims to define a compressed and interoperable representation of trained neural networks. In this paper, we present a low-rank approximation to compress a CNN used for image compression, which is one of MPEG-NNR use cases. In the presented method, the low-rank approximation decomposes one 2D kernel matrix of weights into two 1D kernel matrix values in each convolution layer to reduce the data amount of weights. The evaluation results show that the model size of the original CNN is reduced to half as well as the inference runtime is reduced up to about 30% with negligible loss in PSNR.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.11a
• /
• pp.133-135
• /
• 2020

최근 CNN(Convolutional Neural Network)은 영상 분류, 객체 인식 등 다양한 비전 분야에서 우수한 성능을 보여주고 있으나, CNN 모델의 계산량 및 메모리가 매우 커짐에 따라 모바일 또는 IoT(lnternet of Things) 장치와 같은 저전력 환경에 적용되기에는 제한이 따른다. 따라서, CNN 모델의 임무 성능을 유지하연서 네트워크 모델을 압축하는 기법들이 연구되고 있다. 본 논문에서는 행렬 분해 기술인 저계수행렬 근사(Low-rank approximation)와 CP(Canonical Polyadic) 분해 기법을 결합하여 CNN 모델을 압축하는 기법을 제안한다. 제안하는 기법은 계층의 유형에 상관없이 하나의 행렬분해 기법만을 적용하는 기존의 기법과 달리 압축 성능을 높이기 위하여 CNN의 계층 타입에 따라 두 가지 분해 기법을 선택적으로 적용한다. 제안기법의 성능검증을 위하여 영상 분류 CNN 모델인 VGG-16, ResNet50, 그리고 MobileNetV2 모델 압축에 적용하였고, 모델의 계층 유형에 따라 두 가지의 분해 기법을 선택적으로 적용함으로써 저계수행렬 근사 기법만 적용한 경우 보다 1.5~12.1 배의 동일한 압축율에서 분류 성능이 향상됨을 확인하였다.

• Journal of the Korean Mathematical Society
• /
• v.61 no.3
• /
• pp.427-444
• /
• 2024

The matrix completion problem is to predict missing entries of a data matrix using the low-rank approximation of the observed entries. Typical approaches to matrix completion problem often rely on thresholding the singular values of the data matrix. However, these approaches have some limitations. In particular, a discontinuity is present near the thresholding value, and the thresholding value must be manually selected. To overcome these difficulties, we propose a shrinkage and thresholding function that smoothly thresholds the singular values to obtain more accurate and robust estimation of the data matrix. Furthermore, the proposed function is differentiable so that the thresholding values can be adaptively calculated during the iterations using Stein unbiased risk estimate. The experimental results demonstrate that the proposed algorithm yields a more accurate estimation with a faster execution than other matrix completion algorithms in image inpainting problems.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2024.05a
• /
• pp.600-601
• /
• 2024

본 논문은 RAG(Retrieval-Augmented Generation) End2End의 리소스(Resource) 과부하 문제를 해결하는 동시에 모델 성능을 향상 시키기 위해 PEFT(Parameters-Efficient Fine-Tuning)기술인 LoRA(Low Rank Adaptation)적용에 관한 연구이다. 본 논문에서는 RAG End2End 모델의 파라미터 값과 개수를 유지하면서, LRM(Low Rank Matrices)을 이용하여 추가적인 파라미터만을 미세 조정하는 방식으로, 전반적인 모델의 효율성을 극대화하는 방안을 제시하였다. 본 논문에서 다양한 도메인에 데이터 셋에 대한 제안 방식의 성능을 검증하고자 Conversation, Covid-19, News 데이터 셋을 사용하였다. 실험결과, 훈련에 필요한 파라미터의 크기가 약 6.4억개에서 180만개로 감소하였고, EM(Exact Match)점수가 유사하거나 향상되었다. 이는 LoRA를 통한 접근 법이 RAG End2End 모델의 효율성을 개선할 수 있는 효과적인 전략임을 증명하였다.

• Clean Technology
• /
• v.19 no.4
• /
• pp.464-468
• /
• 2013

In this study, in order to increase solid content of coal water fuel (CWF), various experimental parameters (i.e., coal type, additive, particle size distribution, drying method) were evaluated. To investigate the drying method, specimen is compared to using flash dry, fluidized bed dry and oil deposit stabilized coal. Difference of the solid content between low rank coal and high rank coal in this case indicate that high rank coal exhibits more higher than 20% of the solid cotent. And specimen for dispersibility was prepared by using dispersing agent of 4 types. As a result, using the dispersing agent was shown 5% higher in sold content than the case of not using the dispersing agent. Efficiency of CWF was improved by using fine coal of 80% in the particle size distribution of coal. Result of CWF using drying methods of 3 types, oil deposit stabilized (ODS) coal dried and stabilized was effective 12% higher in sold content than raw coal.

• 한국연소학회:학술대회논문집
• /
• 2014.11a
• /
• pp.289-292
• /
• 2014

• Korean Chemical Engineering Research
• /
• v.50 no.6
• /
• pp.1086-1092
• /
• 2012

In this study, kinetic studies and analysis of the produced syngas were conducted for low rank coal gasification under $CO_2$ atmosphere. 6 coals were analyzed to measure amount of sulfur and ash by proximate and ultimate analyses. And then they were analyzed to select suitable sample by using Thermogravimetric analyzer (TGA). Selected coal sample Samhwa was mixed with catalysts. Mixed samples with catalysts were used to get activation energy under $CO_2$ atmosphere by using Kissinger's method and shrinking core model (SCM). Also, analysis of produced syngas was performed by Gas Chromatography (GC). In this experiment, activation of the $K_2CO_3$ was the best performance, and result of the analysis of the syngas showed similar trend with result of the activation energy.