• Journal of Broadcast Engineering
• /
• v.26 no.6
• /
• pp.778-789
• /
• 2021

Recently, various lightweight methods for using Convolutional Neural Network(CNN) models in mobile devices have emerged. Weight quantization, which lowers bit precision of weights, is a lightweight method that enables a model to be used through integer calculation in a mobile environment where GPU acceleration is unable. Weight quantization has already been used in various models as a lightweight method to reduce computational complexity and model size with a small loss of accuracy. Considering the size of memory and computing speed as well as the storage size of the device and the limited network environment, this paper proposes a method of compressing integer weights after quantization using a video codec as a method. To verify the performance of the proposed method, experiments were conducted on VGG16, Resnet50, and Resnet18 models trained with ImageNet and Places365 datasets. As a result, loss of accuracy less than 2% and high compression efficiency were achieved in various models. In addition, as a result of comparison with similar compression methods, it was verified that the compression efficiency was more than doubled.

• Korean Journal of Soil Science and Fertilizer
• /
• v.46 no.2
• /
• pp.136-141
• /
• 2013

Lowering tree height has been a key practice for efficient management of persimmon orchards. This experiment was conducted to assess whether fertilization rates could be reduced after lowering the tree height by heavy pruning. Twelve-year-old 'Uenishiwase' persimmon trees were lowered to 2 m from over 3 m by severe dormant pruning, and then conventional fertilization rate was applied to some trees in April (N 224 g, $P_2O_5$ 172 g, and $K_2O$ 172 g per tree), June (N 112 g and $K_2O$ 99 g per tree), and October (N 112 g and $K_2O$ 99 g per tree). At the same time, 1/3 or 2/3 of the conventional rate and none were applied to other trees, respectively. Non-application decreased shoot length and reduced number of unnecessary secondary shoots by 39% compared with the conventional rate, not affecting yield and weight, color, firmness, and soluble solids of fruits. No significant difference was also found in the yield and the fruit characteristics among the trees fertilized with different rates. Concentrations of soluble sugars, starch, N, and K of dormant shoots in March of the following year were not significantly changed by the different treatments of the previous year. There was no significant difference of shoot growth and yield among the treatments the following year when the same fertilization rate was supplied to all the trees. Results indicated that fertilization rate could be reduced to less than 1/3 of conventional rate to save the cost and stabilize shoot vigor when tree height is lowered by severe pruning.

• ETRI Journal
• /
• v.27 no.6
• /
• pp.733-746
• /
• 2005

In the context of multi-protocol label switching (MPLS) traffic engineering, this paper proposes a scalable constraintbased shortest path first (CSPF) routing algorithm with multiple QoS metrics. This algorithm, called the multiple constraint-based shortest path first (M_CSPF) algorithm, provides an optimal route for setting up a label switched path (LSP) that meets bandwidth and end-to-end delay constraints. In order to maximize the LSP accommodation probability, we propose a link weight computation algorithm to assign the link weight while taking into account the future traffic load and link interference and adopting the concept of a critical link from the minimum interference routing algorithm. In addition, we propose a bounded order assignment algorithm (BOAA) that assigns the appropriate order to the node and link, taking into account the delay constraint and hop count. In particular, BOAA is designed to achieve fast LSP route computation by pruning any portion of the network topology that exceeds the end-to-end delay constraint in the process of traversing the network topology. To clarify the M_CSPF and the existing CSPF routing algorithms, this paper evaluates them from the perspectives of network resource utilization efficiency, end-to-end quality, LSP rejection probability, and LSP route computation performance under various network topologies and conditions.

• Computers and Concrete
• /
• v.11 no.6
• /
• pp.603-620
• /
• 2013

This paper improves genetic programming (GP) and weight genetic programming (WGP) and proposes soft-computing polynomials (SCP) for accurate prediction and visible polynomials. The proposed genetic programming system (GPS) comprises GP, WGP and SCP. To represent confined compressive strength and strain of circular concrete columns in meaningful representations, this paper conducts sensitivity analysis and applies pruning techniques. Analytical results demonstrate that all proposed models perform well in achieving good accuracy and visible formulas; notably, SCP can model problems in polynomial forms. Finally, concrete compressive strength and lateral steel ratio are identified as important to both confined compressive strength and strain of circular concrete columns. By using the suggested formulas, calculations are more accurate than those of analytical models. Moreover, a formula is applied for confined compressive strength based on current data and achieves accuracy comparable to that of neural networks.

• Proceedings of the Korean Information Science Society Conference
• /
• 2000.10b
• /
• pp.305-307
• /
• 2000

본 논문에서는 다층 퍼셉트론 신경망에서 연결선 수를 최소로 하면서 일반화 성능을 향상시키기 위해 가장 널리 쓰여지고 있는 Optimal Brain Surgeon을 이용한 프루닝(pruning)을 기반으로 하여 오차 함수의 가중치 감소항을 추가시키는 방법을 사용한다. 이때 학습 및 프루닝의 성능에 많은 영향을 미치는 가중치 감소항의 방영정도를 베이시안 테크닉에 기반하여 적응적으로 최적화 하는 방법을 제안한다. 제안하는 방법의 성능을 검증하기 위해 벤치마크 데이터를 이용하여 실험을 수행하였다. 순수한 OBS 방법과 고정된 반영정도를 가진 가중치 감소항을 추가시킨 OBS, 그리고 제안하는 적응적 가중치 감소항을 적용한 OBS 방법을 비교하여 제한하는 방법이 기존의 두 방법에 비해 신경망 구조의 최적화 능력이 뛰어남을 확인할 수 있었다.

• Journal of Korean Society of Forest Science
• /
• v.97 no.3
• /
• pp.221-228
• /
• 2008

This study was conducted to understand the effect of the different treatment level of head back pruning (HBP) on flowering, fruiting and nut qualities. The average number of pistillate flowers per bearing mother branch (BMB) ranged greatly from 0.5 to 60.0. For all cultivars as the level of HBP treatment increases, the number of pistillate flowers per bearing branch (BB) and BMB decreased. For all cultivars the number of branches developed per BMB began to decrease above 30% HBP treatment level. The number of burrs per BB ranged from 1.0 to 2.0. Within 'Daebo' the number of burrs per BB increased with increasing HBP treatment level. The rate of fallen fruits was the lowest of 22.6% at the 50% HBP treatment level of 'Tsukuba', whereas the rate was the highest in 'Pyeonggi'. Nut weight increased with decreasing the number of burrs per BB. Nut yield per BMB showed a tendency of decreasing with increasing the HBP treatment level. Within a HBP treatment level nut grade indicating nut quality varied greatly among cultivars. For all cultivars except for 'Pyeonggi', the proportion of specially large nuts above the average weight increased as the HBP treatment level increased. The soluble solids content showed no significant difference among the HBP treatment levels just right after harvesting of nuts. After 8 weeks of harvesting, however, the soluble solids content increased about 3.3% to 5.3%. and was highest at the 50% HBP treatment level. There were no significant differences among the HBP treatment levels and cultivars in the nut hardness which is highly related to the storage as well as the initial soluble solid content.

• Horticultural Science & Technology
• /
• v.18 no.5
• /
• pp.591-593
• /
• 2000

This study was conducted to investigate the effect of training time on the growth and yield of 'Sinsakigake-2' pepper (Capsicum annuum L.). Plants were either left untrained as control or trained at the third node leaving four shoots per plant. Plants were traind at 20, 45, 70, or 95 days after transplanting (DAT). The earlier the training time, the more the branch number of the plant. Fruit weight was not affected by training time. However, the percentage of marketable fruits and the number of marketable fruits per plant were the highest in plants trained at 70 DAT, and the lowest in the control. Early marketable yield was the highest in the control, and the lowest in plants trained at 45 DAT.

• Journal of the Korea Furniture Society
• /
• v.24 no.4
• /
• pp.345-349
• /
• 2013

Yellow poplar is well known commercial hardwood species in the world because of its availability, rapid growth, large size, excellent form, early natural pruning, and the good working quality of the wood. Therefore, yellow poplar (Liriodendron tulipifera L.) has been adopted in South Korea for decades ago, and its cultivating the saplings and silvicultural system have been studied. However, yellow poplar is still embroiled in controversy for utilization of use and worth. This species has not only advantages such as lower weight, light color, and higher machinability and bending processing, but also has disadvantages which are bad smell and different colors on heartwood and sapwood. In this study, the objective was that use of yellow poplar board as furniture materials by drying and painting process. Thirty years old yellow poplar was cut and dried for 7 days by high-frequency dryness. Adjustment of color tone to walnut color was conducted with multiple colorants and various number of spread times. Uniform color of yellow poplar on heartwood and sapwood was accomplished by applying 1 time stain (green), 2 times transparent primer, 1 time middle-painting (walnut), and 2 times top-painting (walnut) in sequentially. The painted yellow poplar board was used to manufacture table and dressing table. During the furniture manufacture process, no critical issues were occurred on part of jointing and painting. This painting method may be useful data for future use of yellow poplar as furniture materials.

• Journal of Korea Game Society
• /
• v.18 no.5
• /
• pp.59-66
• /
• 2018

In this paper, one of the traditional game, Gonu game, is implemented and experimented. The Minimax algorithm was applied as a technique to implement the Gonu game. We proposed an evaluation function to implement game in Minimax algorithm. We analyze the efficiency of algorithm for alpha beta pruning to improve the performance after implementation of Gonu game. Weights were analyzed for optimal analysis that affected the win or loss of the game. For the weighting analysis, a competition of human and computer was performed. We also experimented with computer and computer. As a result, we proposed a weighting value for optimal attack and defense.

• Journal of KIISE:Databases
• /
• v.36 no.5
• /
• pp.341-351
• /
• 2009

Recently high utility pattern (HUP) mining is one of the most important research issuer in data mining since it can consider the different weight Haloes of items. However, existing mining algorithms suffer from the performance degradation because it cannot easily apply Apriori-principle for pattern mining. In this paper, we introduce new high utility pattern mining approach by using a prefix-tree as in FP-Growth algorithm. Our approach stores the weight value of each item into a node and utilizes them for pruning unnecessary patterns. We compare the performance characteristics of three different prefix-tree structures. By thorough experimentation, we also prove that our approach can give performance improvement to a degree.