• 전자공학회논문지
• /
• 제50권2호
• /
• pp.143-151
• /
• 2013

낸드 플래시 기반 SSD(Solid State Drive)는 빠른 입출력 성능, 저전력 등의 장점을 가지고 있어, 타블릿, 데스크탑 PC, 스마트폰, 서버 등의 저장장치로 널리 사용되고 있다. 하지만 SSD는 쓰기 횟수에 따라서 마모도가 증가하는 단점이 있다. SSD의 수명을 향상시키기 위해 다양한 데이터 중복제거 기법이 도입되었으나, 일반적인 고정 크기 분할방식은 데이터의 지역성을 고려하지 않고 청크크기를 할당함으로써, 불필요한 청킹 및 해시값 생성을 수행하는 문제점이 있으며, 가변 크기 분할방식은 중복제거를 위해 바이트 단위로 비교하여 과도한 연산량을 유발한다. 본 논문에서는 SSD 기반 서버급 스토리지에서 쓰기 요청된 데이터의 지역성에 기반한 적응형 청킹 정책을 제안한다. 제안한 방법은 중복데이터가 가지는 응용프로그램 및 파일 이름 기반 지역성에 따라 청크 크기를 4KB 또는 64KB로 적응적으로 분할하여, 청킹 및 해시값 생성에 따른 오버헤드를 감소시키고, 중복 쓰기를 방지한다. 실험결과, 제안하는 기법이 기존의 가변 크기 분할 및 4KB의 고정 크기 분할을 이용한 중복제거기법보다 SSD의 쓰기 성능이 향상되고 전력 소모 및 연산시간을 감소시킬 수 있음을 보여준다.

• 대한수학회지
• /
• 제54권2호
• /
• pp.613-626
• /
• 2017

For the overdetermined linear system, when both the data matrix and the observed data are contaminated by noise, Total Least Squares method is an appropriate approach. Since an ill-conditioned data matrix with noise causes a large perturbation in the solution, some kind of regularization technique is required to filter out such noise. In this paper, we consider a Dual regularized Total Least Squares problem. Unlike the Tikhonov regularization which constrains the size of the solution, a Dual regularized Total Least Squares problem considers two constraints; one constrains the size of the error in the data matrix, the other constrains the size of the error in the observed data. Our method derives two nonlinear equations to construct the iterative method. However, since the Jacobian matrix of two nonlinear equations is not guaranteed to be nonsingular, we adopt a trust-region based iteration method to obtain the solution.

• 한국통신학회논문지
• /
• 제22권11호
• /
• pp.2401-2408
• /
• 1997

The measured pulse shapes provided in the given data package demonstrated pulse distortions due to laser speckle. the distorted pulse shapes were carefully analyzed, modeled, and then applied to the DTTL(Digital-data Transition Tracking Loop)[1] bit synchronizer simulator to measure the mean time delay and its delay variation performance. The result showed that the maximum mean time delay variation with the modeled data was 12.5% when window size equals 1. All the data given were located within this modeled boundary and the maximum eman time delay variation was 7% in this case. The mean time delay variation was known to be smaller by reducing the window size [2][5][6]. The mitigated delay variation was 2.5% in the modeled case and 1.4% in the data set given when the windown size equals 0.1. With the digital DTTL insteal of analog DTTL, similar results was obtained.

• 대한임베디드공학회논문지
• /
• 제6권6호
• /
• pp.383-391
• /
• 2011

As we know the effects of cache memory research, instruction and data caches can be separated for higher performance with Harvard CPUs. In this paper, we shows the efficiency of buffer system in the instruction and data flash storage medium. And we analyzed characteristics of the data and instruction flash and evaluated the performance. Finally, we propose the best buffer structure with an optimal block size and buffer size for the instruction and data flash.

• The Journal of Korean Physical Therapy
• /
• 제32권5호
• /
• pp.277-282
• /
• 2020

Purpose: The main purposes of this study are to identify the degree of effect size and variables related to it, on the effect of physical therapy on shoulder adhesive capsulitis. Methods: This study were collected 8 studies published between 2010.01.01. and 2020.08.01. The Analysis result confirmed 16 effect size data, the random effect model was chosen because of the heterogeneity of the data. Results: First, full case showed the largest mean effect size 3.28 (p<0.001). Second, Size of the effect according to hands on+off intervention the 5.17 (p>0.344). Third, Number of participants showed 15 or more effect size 4.19 (p<0.002). intervention period showed 8 weeks effect size 5.12 (p>0.425). Number of intervention showed 14 or less effect size 3.99 (p>0.061). Publication type showed thesis effect size 4.34 (p<0.003). Finally, 'Trim and Fill' result confirmed that the calibration effect size is 1.71 (p<0.001). Conclusion: We were able to confirm that physical therapy had great effect on shoulders adhesive capsulitis, and that the effect differed according to the methods of intervention, the methods of evaluation.

• 한국의류산업학회지
• /
• 제15권2호
• /
• pp.247-255
• /
• 2013

This study develops size sections and specifications for females 7-18 years old. The specifications will contribute to the development of precise virtual models that represent basic data for the size alteration of 3D virtual models. A total of 33 items from the $5^{th}$ Size Korea data set were adopted and analyzed. Two representative factors (Obesity and Height Factors) were identified through a factor analysis of the 33 items. Waist circumference and stature were selected as representative variables for Obesity and Height Factors, respectively, and conducted cross tabulations between waist circumference and height for the age groups of 7-12 and 13-18 year-olds. Size sections were developed for the development of 3D models based on the results. A regression analysis then developed size specifications for each section with independent variables of waist circumference and height as well as dependent variables of reference body size. Subsequently, Obesity Factors were better explained by waist circumference and Height Factors were better explained by stature. Finally, size specifications for each section were developed.

• The Journal of Korean Physical Therapy
• /
• 제32권4호
• /
• pp.222-227
• /
• 2020

Purpose: The primary purposes of this study were to identify the degree of the effect size and the variables related to it on the effects of physical therapy on myofascial pain syndrome. Methods: This study collected 15 studies published between 2008.01.01 and 2019.12.31. The Analysis results confirmed 57 effect size data. The random-effect model was chosen because of the heterogeneity of the data. Results: First, the full case showed the largest mean effect size of 2.03 (p<0.001). Second, the size of the effect, according to the handson intervention, was 2.74 (p<0.003). Third, the VAS showed an effect size of 2.30 (p<0.001). Fourth, the intervention period showed a 1- to 15-day effect size of 2.94 (p<0.001). The number of interventions showed a 6 to 10 effect size of 2.84 (p<0.006). The number of participants showed a 10 or less effect size of 2.66 (p<0.001). Finally, the 'Trim and Fill' result confirmed that the calibration effect size was 0.93 (p<0.001). Conclusion: Physical therapy had a great effect on myofascial pain syndrome in the neck and shoulders, and that the effect differed according to the methods of the intervention and the methods of evaluation.

• 한국정보과학회논문지:정보통신
• /
• 제30권2호
• /
• pp.155-166
• /
• 2003

이동 컴퓨팅은 최근 이동 통신 장비의 발전과 이동 통신 환경의 인터넷 연동으로 인해 컴퓨팅 연구 분야의 새로운 이슈로 등장했다. 이동 통신 환경은 잦은 접속 단절, 한정된 대역폭, 이동 클라이언트의 에너지 용량 제한 등 많은 제약이 있다. 이러한 제약으로 인해 많은 응용에서 사용자에게 데이타를 효율적으로 전달하기 위해 일반적으로 브로드캐스팅을 사용한다. 본 논문에서는 인기도와 데이타 크기를 고려하여 브로드캐스트하는 주기적 브로드캐스팅 채널과 이동 클라이언트의 요구에 따라 브로드캐스트하는 주문형 브로드캐스팅 채널로 브로드캐스팅 채널을 논리적으로 분할하고, 각 채널의 대역폭을 동적으로 할당하는 적응적 브로드캐스팅 기법을 제안한다. 제안하는 브로드캐스팅 기법은 브로드캐스트될 데이터를 선택할 패, 이동 클라이언트의 이동성을 고려하며, 본 논문에서 고안된 크기 요소SF(Size Factor)를 사용하여 대상 데이타의 크기를 고려한다. 본 논문에서는 또한 실험을 통해 이동 클라이언트의 전력 소모량을 측정하여 제안하는 브로드캐스팅 기법의 성능을 평가한다.

• 한국정보통신학회논문지
• /
• 제26권1호
• /
• pp.49-57
• /
• 2022

최근 인공지능의 다양한 활용은 기계학습의 딥 인공신경망 구조를 통해 가능해졌으며 인간과 같은 능력을 보여주고 있다. 불행하게도 딥 구조의 인공신경망은 아직 정확한 해석이 이루어지고 있지 못하고 있다. 이러한 부분은 인공지능에 대한 불안감과 거부감으로 작용하고 있다. 우리는 이러한 문제 중에서 인공신경망의 능력 부분을 해결한다. 인공신경망 구조의 크기를 계산하고, 그 인공신경망이 처리할 수 있는 데이터의 크기를 계산해 본다. 계산의 방법은 수학에서 쓰이는 군의 방법을 사용하여 데이터와 인공신경망의 크기를 군의 구조와 크기를 알 수 있는 Order를 이용하여 계산한다. 이를 통하여 인공신경망의 능력을 알 수 있으며, 인공지능에 대한 불안감을 해소할 수 있다. 수치적 실험을 통하여 데이터의 크기와 딥 인공신경망을 계산하고 이를 검증한다.

• 한국측량학회:학술대회논문집
• /
• 한국측량학회 2007년도 춘계학술발표회 논문집
• /
• pp.187-190
• /
• 2007

This study is for analyzing best band in image matching using correlation coefficient of left and right images of stereo image pair, lot red, green, blue band images separated from color aerial photo and gray image converted from the same color aerial photo image. The image matching is applied to construct Digital Elevation Model(DEM) or terrain data. The correlation coefficients and variation by change of pixel patch size are computed from pixel patches of which sizes are $11{\times}11{\sim}101{\times}101$. Consequently, the correlation coefficient in red band image is highest. The lowest is in blue band. Therefore, to construct terrain data using image matching, the red band image is preferable. As the size of pixel patch is growing, the correlation coefficient is increasing. But increasing rate declines from $51{\times}51$ image patch size and above. It is proved that the smaller pixel patch size than $51{\times}51$ is applied to construct terrain data using image matching.