• Journal of the Institute of Convergence Signal Processing
• /
• v.22 no.3
• /
• pp.99-103
• /
• 2021

Recently, convolutional neural networks (CNNs) have been widely utilized due to their outstanding performance in various computer vision fields. However, due to their computational-intensive and high memory requirements, it is difficult to deploy CNNs on hardware platforms that have limited resources, such as mobile devices and IoT devices. To address these limitations, a neural network compression research is underway to reduce the size of neural networks while maintaining their performance. This paper proposes a CNN compression technique that dynamically adjusts the thresholds of pruning, one of the neural network compression techniques. Unlike the conventional pruning that experimentally or heuristically sets the thresholds that determine the weights to be pruned, the proposed technique can dynamically find the optimal thresholds that prevent accuracy degradation and output the light-weight neural network in less time. To validate the performance of the proposed technique, the LeNet was trained using the MNIST dataset and the light-weight LeNet could be automatically obtained 1.3 to 3 times faster without loss of accuracy.

• Structural Engineering and Mechanics
• /
• v.78 no.2
• /
• pp.125-134
• /
• 2021

In this paper, the effects of Tuned Mass dampers (TMDs) on the reduction of the vertical vibrations of a real horizontally curved steel box-girder bridge due to different traffic loads are numerically investigated. The performance of TMDs to reduce the bridge vibrations can be affected by the parameters such as dynamic characteristics of TMDs, the location of TMDs, the speed and weight of vehicles. In the first part of this study, the effects of mass ratio, damping percentage, frequency ratio, and location of TMDs on the performance of TMDs to decrease vertical vibrations of different sections of bridge deck are evaluated. In the second part, the performance of TMD is investigated for different speeds and weights of traffic loads. Results show that the mass ratio of TMDs is the more effective parameter in reducing imposed vertical vibration in comparison with the damping ratio. Furthermore, it is found that TMD is very sensitive to its tuned frequency, i.e., with a little deviation from a suitable frequency, the expected performance of TMD significantly decreased. TMDs have a positive and considerable performance at certain vehicle speeds and this performance declines when the weight of traffic loads is increased. Besides, the results reveal that the highest impact of TMD on the reduction of the vertical vibrations is when free vibrations occur for the bridge deck. In that case, maximum reductions of 24% and 59% are reported in the vertical acceleration of the bridge deck for the forced and free vibration amplitudes, respectively. The maximum reduction of 13% is also obtained for the maximum displacement of the bridge deck. The results are mainly related to the resonance condition.

• Animal Bioscience
• /
• v.36 no.11
• /
• pp.1727-1737
• /
• 2023

Objective: The poultry industry is a primary source of animal protein worldwide. The gut microbiota of poultry birds, such as chickens and ducks, is critical in maintaining their health, growth, and productivity. This study aimed to identify longitudinal changes in the gut microbiota of laying hens from birth to the pre-laying stage. Methods: From a total of 80 Hy-Line Brown laying hens, birds were selected based on weight at equal intervals to collect feces (n = 20 per growth) and ileal contents (n = 10 per growth) for each growth stage (days 10, 21, 58, and 101). The V4 regions of the 16S rRNA gene were amplified after extracting DNA from feces and ileal contents. Amplicon sequencing was performed using Illumina, followed by analysis. Results: Microbial diversity increased with growth stages, regardless of sampling sites. Microbial community analysis indicated that Firmicutes, Proteobacteria, and Bacteroidetes were the dominant phyla in the feces and ileal. The abundance of Lactobacillus was highest on day 10, and that of Escherichia-shigella was higher on day 21 than those at the other stages at the genus level (for the feces and ileal contents; p<0.05). Furthermore, Turicibacter was the most abundant genus after changing feed (for the feces and ileal contents; p<0.05). The fecal Ruminococcus torques and ileal Lysinibacillus were negatively correlated with the body weights of chickens (p<0.05). Conclusion: The gut microbiota of laying hens changes during the four growth stages, and interactions between microbiota and feed may be present. Our findings provide valuable data for understanding the gut microbiota of laying hens at various growth stages and future applied studies.

• Korean Journal of Agricultural Science
• /
• v.24 no.2
• /
• pp.207-217
• /
• 1997

This study was performed to evaluate the engineering properties of rice-husk ash concrete using normal portland cement, natural aggregates and rice-husk ash. The following conclusions were drawn; 1. The unit weight was in the range of $2,216{\sim}2,325kgf/m^3$, the weights of those concrete were decreased 1~6% than that of the normal cement concrete, respectively. 2. The highest strength was achieved by 10% rice-husk ash filled rice-husk ash concrete, it was increased 8% by compressive strength, 17% by tensile strength and 18% by bending strength than that of the normal cement concrete, respectively. 3. The ultrasonic pulse velocity was in the range of 3,252~4,016 m/s, which was showed about the same compared to that of the normal cement concrete. The highest ultrasonic pulse velocity was showed by 10% rice-husk ash filled rice-husk ash concrete. 4. The dynamic modulus of elasticity was in the range of $242{\times}10^3{\sim}306{\times}10^3kgf/cm^2$, which was showed about the same compared to that of the normal cement concrete. The highest dynamic modulus was showed by 10% rice-husk ash filled rice-husk ash concrete. 5. The static modulus of elasticity was in the range of $185{\times}10^3{\sim}275{\times}10^3kgf/cm^2$, which was showed about the same compared to that of the normal cement concrete. The poisson's number of rice-husk ash concrete was less than that of the normal cement concrete. The dynamic modulus was increased approximately 11~30% than that of the static modulus. 6. The durability was increased with increase of the content of rice-husk ash. The durability was increased 1.3 times by 10% rice-husk ash, 1.6times by 20% rice-husk ash filled concrete than that of the normal cement concrete. respectively.

• Journal of KIISE:Computer Systems and Theory
• /
• v.34 no.10
• /
• pp.529-538
• /
• 2007

Fair Queueing algorithms based on Generalized Processor Sharing (GPS) not only guarantee sessions with service rate and delay, but also provide sessions with instantaneous fair sharing. This fair sharing distributes server capacity to currently backlogged sessions in proportion to their weights without regard to the amount of service that the sessions received in the past. From a long-term perspective, the instantaneous fair sharing leads to a different quality of service in terms of delay and bandwidth to sessions with the same weight depending on their traffic pattern. To minimize such long-term unfairness, we propose a delay-bandwidth normalization model that defines the concept of value of service (VoS) from the aspect of both delay and bandwidth. A model and a packet-by-packet scheduling algorithm are proposed to realize the VoS concept. Performance comparisons between the proposed algorithm and algorithms based on fair queueing and service curve show that the proposed algorithm provides better long-term fairness among sessions and that is more adaptive to dynamic traffic characteristics without compromising its service rate and delay guarantees.

• KSBB Journal
• /
• v.19 no.4
• /
• pp.274-283
• /
• 2004

Six lumbrokinase (LK) fractions from Lumbricus rubellus lysates were purified by a series of column chromatographies. The molecular weights of the six LK fractions appeared to range from 24.6 to 33.1 kDa. In the experimental model of rat venous thrombosis, the thrombus weight and PAI activity decreased significantly when the LK was administered orally. However, the activities of APTT, PT and plasmin showed a significant increase. The aggregation of rat platelets pretreated with various LK doses was inhibited by thrombin, and the MDA generation decreased. The rat thoracic aorta and mesentric arteries contracted with phenylephrine relaxed due to the treatment of the LK fractions. These results suggest that the fibrinolytic effects of LK were mediated not only by proteolytic activity, but also by the inhibition of platelet agregation and the relaxation of blood vessels. It is concluded that the LK may be useful as a hemolytic agent for treatment of fibrin clot.

• The KIPS Transactions:PartB
• /
• v.15B no.5
• /
• pp.457-464
• /
• 2008

In the real-world operational environment, most of text classification systems have the problems of insufficient training documents and no prior knowledge of feature space. In this regard, $Na{\ddot{i}ve$ Bayes is known to be an appropriate algorithm of operational text classification since the classification model can be evolved easily by incrementally updating its pre-learned classification model and feature space. This paper proposes the improving technique of $Na{\ddot{i}ve$ Bayes classifier through feature weighting strategy. The basic idea is that parameter estimation of $Na{\ddot{i}ve$ Bayes considers the degree of feature importance as well as feature distribution. We can develop a more accurate classification model by incorporating feature weights into Naive Bayes learning algorithm, not performing a learning process with a reduced feature set. In addition, we have extended a conventional feature update algorithm for incremental feature weighting in a dynamic operational environment. To evaluate the proposed method, we perform the experiments using the various document collections, and show that the traditional $Na{\ddot{i}ve$ Bayes classifier can be significantly improved by the proposed technique.

• Computers and Concrete
• /
• v.22 no.3
• /
• pp.337-353
• /
• 2018

The present study focuses on examining the structural behaviour of steel-fibre-reinforced concrete (SFRC) beams under high rates of loading largely associated with impact problems. Fibres are added to the concrete mix to enhance ductility and energy absorption, which is important for impact-resistant design. A simple, yet practical non-linear finite-element analysis (NLFEA) model was used in the present study. Experimental static and impact tests were also carried out on beams spanning 1.3 meter with weights dropped from heights of 1.5 m and 2.5 m, respectively. The numerical model realistically describes the fully-brittle tensile behaviour of plain concrete as well as the contribution of steel fibres to the post-cracking response (the latter was allowed for by conveniently adjusting the constitutive relations for plain concrete, mainly in uniaxial tension). Suitable material relations (describing compression, tension and shear) were selected for SFRC and incorporated into ABAQUS software Brittle Cracking concrete model. A more complex model (i.e., the Damaged Plasticity concrete model in ABAQUS) was also considered and it was found that the seemingly simple (but fundamental) Brittle Cracking model yielded reliable results. Published data obtained from drop-weight experimental tests on RC and SFRC beams indicates that there is an increase in the maximum load recorded (compared to the corresponding static one) and a reduction in the portion of the beam span reacting to the impact load. However, there is considerable scatter and the specimens were often tested to complete destruction and thus yielding post-failure characteristics of little design value and making it difficult to pinpoint the actual load-carrying capacity and identify the associated true ultimate limit state (ULS). To address this, dynamic NLFEA was employed and the impact load applied was reduced gradually and applied in pulses to pinpoint the actual failure point. Different case studies were considered covering impact loading responses at both the material and structural levels as well as comparisons between RC and SFRC specimens. Steel fibres were found to increase the load-carrying capacity and deformability by offering better control over the cracking process concrete undergoes and allowing the impact energy to be absorbed more effectively compared to conventional RC members. This is useful for impact-resistant design of SFRC beams.

• Journal of Korean Society of Transportation
• /
• v.23 no.5 s.83
• /
• pp.135-146
• /
• 2005

Up to now studies on the fusion of travel time from various detectors have been conducted based on the variance raito of the intermittent data mainly collected by GPS or probe vehicles. The fusion model based on the variance ratio of intermittent data is not suitable for the license plate recognition AVIs which can deal with vast amount of data. This study was carried out to develop the fusion model based on travel time acquired from the license plate recognition AVIs and the point detectors. In order to fuse travel time acquired from the point detectors and the license plate recognition AVIs, the optimized fusion model and the proportional fusion model were developed in this study. As a result of verification, the optimized fusion model showed the superior estimation performance. The optimized fusion model is the dynamic fusion ratio estimation model on real time base, which calculates fusion weights based on real time historic data and applies them to the current time period. The results of this study are expected to be used effectively for National Highway Traffic Management System to provide traffic information in the future. However, there should be further studies on the Proper distance for the establishment of the AVIs and the license plate matching rate according to the lanes for AVIs to be established.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.42 no.1
• /
• pp.9-26
• /
• 2005

In the pursuit of ever higher levels of performance, recent computer systems have made use of deep pipeline, dynamic scheduling and multi-issue superscalar processor technologies. In this situations, branch prediction schemes are an essential part of modem microarchitectures because the penalty for a branch misprediction increases as pipelines deepen and the number of instructions issued per cycle increases. In this paper, we propose a novel branch prediction scheme, direction-gshare(d-gshare), to improve the prediction accuracy. At first, we model a neural network with the components that possibly affect the branch prediction accuracy, and analyze the variation of their weights based on the neural network information. Then, we newly add the component that has a high weight value to an original gshare scheme. We simulate our branch prediction scheme using Simple Scalar, a powerful event-driven simulator, and analyze the simulation results. Our results show that, compared to bimodal, two-level adaptive and gshare predictor, direction-gshare predictor(d-gshare. 3) outperforms, without additional hardware costs, by up to 4.1% and 1.5% in average for the default mont of embedded direction, and 11.8% in maximum and 3.7% in average for the optimal one.