• 로봇학회논문지
• /
• 제17권2호
• /
• pp.133-141
• /
• 2022

This paper introduces model compression algorithms which make a deep neural network smaller and faster for embedded systems. The model compression algorithms can be largely categorized into pruning, quantization and knowledge distillation. In this study, gradual pruning, quantization aware training, and knowledge distillation which learns the activation boundary in the hidden layer of the teacher neural network are integrated. As a large deep neural network is compressed and accelerated by these algorithms, embedded computing boards can run the deep neural network much faster with less memory usage while preserving the reasonable accuracy. To evaluate the performance of the compressed neural networks, we evaluate the size, latency and accuracy of the deep neural network, DenseNet201, for image classification with CIFAR-10 dataset on the NVIDIA Jetson Xavier.

• International journal of advanced smart convergence
• /
• 제11권1호
• /
• pp.28-35
• /
• 2022

This paper presents a performance evaluation of real-time Linux for industrial real-time platforms. On industrial platforms, multicore processors are popular due to their work distribution efficiency and cost-effectiveness. Multicore processors, however, are not designed for applications with real-time constraints, and their performance capabilities depend on their core configurations. In order to assess the feasibility of a multicore processor for real-time applications, we conduct a performance evaluation of a general processor and a low-power processor to provide an experimental environment of real-time Linux on both Xenomai and RT-preempt considering the multicore configuration. The real-time performance is evaluated through scheduling latency and in an environment with loads on the CPU, memory, and network to consider an actual situation. The results show a difference between a low-power and a general-purpose processor, but from developer's point of view, it shows that the low-power processor is a proper solution to accommodate low power situations.

• Journal of Information Processing Systems
• /
• 제18권6호
• /
• pp.741-754
• /
• 2022

Openstack is widely used as a representative open-source infrastructure of the service (IaaS) platform. The Openstack Identity Service is a centralized approach component based on the token including the Memcached for cache, which is the in-memory key-value store. Token validation requests are concentrated on the centralized server as the number of differently encrypted tokens increases. This paper proposes the practical Byzantine fault tolerance (PBFT) blockchain-based Openstack Identity Service, which can improve the performance efficiency and reduce security vulnerabilities through a PBFT blockchain framework-based decentralized approach. The experiment conducted by using the Apache JMeter demonstrated that latency was improved by more than 33.99% and 72.57% in the PBFT blockchain-based Openstack Identity Service, compared to the Openstack Identity Service, for 500 and 1,000 differently encrypted tokens, respectively.

• 약학회지
• /
• 제57권5호
• /
• pp.323-329
• /
• 2013

This study was designed to determine and compare the cognitive enhancing effect of various Codonopsis lanceolata extracts by steaming, fermentation and high hydrostatic pressure process. We prepared water extract of C. lanceolata, steamed C. lanceolata, steamed and fermented C. lanceolata and C.lanceolata by high hydrostatic pressure process and fermentation. Cognitive enhancing effect of extracts was evaluated in scopolamine-induced memory impairment mice using by passive avoidance test and Morris water maze tests. MTT assay was conducted to investigate neuroprotective effect on glutamate induced cell death in HT22 cells. Steamed and fermented C. lanceolata water extract decreased escape latency in Morris water maze test and increased the latency time of the passive avoidance test compared to other extracts. Furthermore, the steamed and fermented C. lanceolata water extract showed neuroprotective effect. These results suggest that steaming and fermentation process more improve cognitive enhancing effect of C. lanceolata than other process.

• 동의신경정신과학회지
• /
• 제18권3호
• /
• pp.55-82
• /
• 2007

Ohjective: This research investigates the effect of the DJR on Alzheimer's disease. Method: 1.The effects of the DJR extract on IL.-$1{\beta}$, IL-6, TNF-${\alpha}$, cox-2, and NOS-II mRNA of BV2 microglia cell line treated with LPS; 2. the behavior: 3. the infarction area of the hippocampus, and brain tissue injury in Alzheimer's diseased mice induced with ${\beta}$A were investigated. Result: 1. The DJR extract suppressed the expression of IL-$1{\beta}$, IL-6 and TNF-${\alpha}$ mRNA in BV2 microglia cell line treated with LPS. 2. The DJR extract suppressed the expression of IL-$1{\beta}$, IL-6, and TNF-${\alpha}$ protein production in BV2 microglia cell line treated with LPS. 3. For the DJR extract group a significant inhibitory effect on the memory deficit was shown for the mice with Alzheimer's disease induced by .${\beta}$A in the Moms water maze experiment, which measured stop-through latency, and distance movement-through latency. 4. The DJR extract suppressed the over-expression of IL-$1{\beta}$ protein, TNF-${\alpha}$ protein and CD68/CD11b, in the mice with Alzheimer's disease induced by ${\beta}$A 5. The DJR extract reduced the infarction area of hippocampus, and controlled the injury of brain tissue in the mice with Alzheimer's disease induced by ${\beta}$A. 6. The DJR extract reduced the tau protein, GFAP protein, and presenilin1/2 protein (immunohistochemistry) of hippocampus in the mice with Alzheimer's disease induced by ${\beta}$A. Conclusion: These results suggest that the DJR extract may he effective for the prevention and treatment of Alzheimer's disease. Investigation into the clinical use of the DJR extract for Alzheimer's disease of suggested for future research.

• Natural Product Sciences
• /
• 제22권4호
• /
• pp.299-306
• /
• 2016

This study aimed to establish the quantitative method to analyze the content of peroxynitrite-scavengers belonging to polyphenols in six Korean Quercus species (Quercus mongolica, Q. dentata, Q. acutissima, Q. alienta, Q. serrata, and Q. variabilis) by HPLC. The twelve peroxynitrite-scavengers, flavanols (catechins: (+)-catechin, (-)-epicatechin, and (-)-epigallocatechin), flavonols (kaempferol and quercetin), flavonol glycosides (astragalin, quercitrin, and isoquercitrin), flavonol acylated glycosides (astragalin 6''-gallate and isoquercitrin 6''-gallate), gallic acid and its dimer (ellagic acid) were analyzed by HPLC. Further, anti-Alzheimer's activity was assayed in a passive avoidance testusing mice by measuring the retention latency (sec), the concentration of acetylcholine (ACh), and acetylcholinesterase (AChE) activity. Simultaneous analysis of the extracts of the six Quercus leaves was achieved on a Capcell C18 column ($5{\mu}m$, $250mm{\times}4.6mm\;i.d.$) with a gradient elution of 0.05% HAc and 0.05% HAc in $CH_3CN$. In the extract of Q. mongolica leaves, the content of gallic acid (32.53 mg/g), (+)-catechin (28.78 mg/g), (-)-epicatehin (22.03 mg/g), astragalin 6''-gallate (20.94 mg/g), and isoquercitrin 6''-gallate (44.11 mg/g) and peroxynitrite-scavenging activity ($IC_{50}$, $0.831{\mu}g/ml$) were high. This extract delayed the retention latency and inhibited acetylcholinesterase activity in scopolamine-induced memory impairment of mice, suggesting that it has anti-Alzheimer's activity.

• 한국콘텐츠학회논문지
• /
• 제18권3호
• /
• pp.34-40
• /
• 2018

본 연구에서는 모바일 단말에서 재생되는 동영상 서비스의 끊김 현상을 방지할 수 있는 최적의 버퍼 크기를 산정하고자 한다. 버퍼는 VOD서비스를 위해 수신 단말에 도착한 멀티미디어 패킷의 저장 공간을 의미한다. 만약 버퍼 크기가 너무 크면, 재생 서비스 시작 전의 지연시간이 길어지게 된다. 반대로 버퍼 크기가 너무 작은 경우, 수신 단말에 도착한 패킷의 부족으로 재생 서비스가 중간에 멈출 수 있다. 그러므로 적정 버퍼 크기는 모바일 단말에서의 재생 서비스 품질 척도가 될 수 있다. 본 연구에서는 재생 서비스의 버퍼링 과정을 이산시간 대기행렬로 모델링한다. N정책을 갖는 Geo/G/1 대기행렬에서 바쁜기간의 평균길이와 평균 대기시간을 분석한다. 또한 도출된 성능척도를 이용하여 모바일 단말에서의 최적의 버퍼 크기를 결정하는 수치예제를 제시한다. 본 연구 결과를 이용하여 재생 도중의 끊김 현상 방지와 재생 시작 초기의 지연시간 최소화를 달성하여 이용자 만족도를 높일 수 있을 것이다.

• 대한전자공학회논문지SD
• /
• 제47권11호
• /
• pp.54-64
• /
• 2010

Reed-Solomon(RS) 코드는 강력한 에러 정정 능력으로 널리 사용된다. 최근 Sudan에 의해 Reed-Solomon 코드의 리스트 디코딩 알고리즘이 정립되었다. 리스트 디코더는 일반적인 디코더보다 더 큰 디코딩 반경을 가지며 하나 이상의 코드를 찾아낸다. 리스트 디코더는 복잡도와 latency가 매우 큰 Interpolation 과 Factorization 단계를 포함하므로 효율적인 하드웨어 설계가 필요하다. Factorization 은 latency가 매 단계마다 변하는 특성을 가져 복잡도가 높으며, 하드웨어 효율 저하의 문제가 발생한다. 본 논문에서는 하드웨어의 재사용을 높인 구조와 알고리즘의 효율적인 처리 스케쥴을 제안한다. 제안한 구조는 각 단계를 작은 단위의 R-MAC 유닛으로 나누어 매 단계마다 하드웨어를 재구성하여 처리함으로서 높은 하드웨어 효율과 효율적인 메모리 구조를 통해 복잡도가 낮은 순차처리를 적용하면서도 높은 처리량을 보이며, 여러 가지 어플리케이션에 적용가능하다. 제안한 구조는 동부 아남 $0.18{\mu}m$ 표준 셀 라이브러리를 사용하여 합성한 결과 최대 동작 주파수는 330MHz이다.

• 전자공학회논문지
• /
• 제49권12호
• /
• pp.149-155
• /
• 2012

SSD(Solid State Disk)는 다수의 NAND 플래시 메모리로 구성되었으며 내부에 고성능 컨트롤러와 캐시 버퍼를 포함한 스토리지 장치이다. NAND 플래시 메모리는 제자리 덮어쓰기가 안되기 때문에 파일시스템에서 유효페이지가 갱신 및 삭제시 무효페이지로 전환되어 완전히 삭제하기 위해서는 가비지 컬렉션 과정을 거쳐야한다. 하지만 가비지 컬렉션은 지연시간이 긴 Erase 연산을 포함하기 때문에 SSD의 I/O 성능을 감소시키고 마모도를 증가시키는 문제가 된다. 본 논문에서는 입력데이터에 대하여 유효데이터와 무효데이터에서 중복검사를 실행하는 기법을 제안한다. 먼저 유효데이터에 대한 중복제거 과정을 거치고 그 다음에 무효데이터 재활용 과정을 거침으로써 중복률을 향상시켰다. 이를 통하여 SSD의 쓰기 횟수와 가비지 컬렉션 횟수를 감소시켜 마모도와 I/O 성능이 개선되었다. 실험결과 제안한 기법은 유효데이터 중복제거와 무효데이터 재활용을 둘다 하지 않는 일반적인 경우에 비해서 가비지 컬렉션 횟수가 최대 20% 감소하고 I/O 지연시간이 9% 감소하였다.

• 동의신경정신과학회지
• /
• 제17권1호
• /
• pp.59-78
• /
• 2006

Objective : This research investigates the effect of the CMT and SCMT on Alzheimer's disease. Methods : The effects of the CMT and SCMT on (1) amyloid precursor proteins(APP), acetylcholinesterase(AChE) mRNA of PC-12 cells treated with CT-105; (2) the AChE activity and the APP production of PC-12 cell treated with CT-105; (3) the behavior; (4) expression of $IL-1{\beta}$, $TNF-{\alpha}$, MDA; (5) the infarction area of the hippocampus in Alzheimer's disease mice induced with CT105 & ${\beta}A$ were investigated Rresults : 1. The CMT and SCMT suppressed the expression of APP, AChE mRNA in PC-12 cells treated with CT-105 2. The CMT and SCMT suppressed the AChE activity, and the production of APP significantly in PC-12 cells treated with CT-105. 3. For the CMT and SCMT group a significant inhibitory effect on the memory deficit was shown for the mice with Alzheimer's disease induced by ${\beta}A$ in the Morris water maze experiment, which measured stop-through latency, and distance movement-through latency 4. The CMT and SCMT suppressed the over-expression of $IL-1{\beta}$, $TNF-{\alpha}$, MDA in the mice with Alzheimer's disease induced by ${\beta}A$. 5. The CMT and SCMT reduced the infarction area of hippocampus with Alzheimer's disease induced by ${\beta}A$. Conclusions : These results suggest that the CMT and SCMT may be effective for the prevention and treatment of Alzheimer's disease. Investigation into the clinical use of the CMT and SCMT for Alzheimer's disease is suggested for future research.