• KIPS Transactions on Computer and Communication Systems
• /
• v.10 no.7
• /
• pp.191-198
• /
• 2021

Since the size of training dataset become large and the model is getting deeper to achieve high accuracy in deep learning, the deep neural network training requires a lot of computation and it takes too much time with a single node. Therefore, distributed deep learning is proposed to reduce the training time by distributing computation across multiple nodes. In this study, we propose hybrid allreduce strategy that considers the characteristics of each layer and communication and computational overlapping technique for synchronization of distributed deep learning. Since the convolution layer has fewer parameters than the fully-connected layer as well as it is located at the upper, only short overlapping time is allowed. Thus, butterfly allreduce is used to synchronize the convolution layer. On the other hand, fully-connecter layer is synchronized using ring all-reduce. The empirical experiment results on PyTorch with our proposed scheme shows that the proposed method reduced the training time by up to 33% compared to the baseline PyTorch.

• Journal of Bio-Environment Control
• /
• v.28 no.1
• /
• pp.16-27
• /
• 2019

Using virtual reality technology, users can learn and experience many interactions in virtual space like the actual physical space. This study was conducted to develop air flow simulator that allows farmers and consultants to consult air flow through VR devices by creating a greenhouse or pigpen model. It can help educate farmers about the importance of ventilation effects for agricultural facilities. We proposed CFD visualization system by building a virtual reality environment and constructing database of CFD and structure of agricultural facilities. After consultants can set up situations according to environmental conditions, the users experience the visualized air flow of agricultural facility according to the ventilation effects. Also it can provide a quantified environmental distribution in the agricultural facility. Currently, the CFD data in agricultural facilities are established during winter and summer. In order to experience various environmental conditions in the developed system, The experts need to run CFD data under various environmental conditions and register them in the system requirements.

• Journal of the Society of Naval Architects of Korea
• /
• v.56 no.1
• /
• pp.34-46
• /
• 2019

This paper provides the numerical results for the self-propulsion performance in waves of a car ferry vessel with damage in one of its twin-screw propulsion systems without flooding the engine room. The numerical simulations were carried out according to the Safe Return to Port (SRtP) regulation made by the Lloyd's register, where the regulation requires that damaged passenger ships should have an ability to return to port with a speed of 6 knots in a Beaufort 8 sea condition. For the validation of the present numerical analysis study, the resistance performance and the self-propulsion performance of the car ferry in intact and damaged conditions in calm water were calculated, which showed a satisfactory agreement with the model test results of Korea Research Institute of Ship and Ocean engineering (KRISO). Finally, the numerical simulation of self-propulsion performance in waves of the damaged car ferry ship was carried out for a normal sea state and for a Beaufort 8 sea state, respectively. The estimated average Brake Horse Power (BHP) for keeping the damaged car ferry ship advancing at a speed of 6 knots in a Beaufort 8 sea state reached about 47% of BHP at MCR condition or about 56% of BHP at NCR condition of the engine determined at the design state. In conclusion, it can be noted that the engine power of the damaged car ferry ship in single propulsion condition is sufficient to satisfy the SRtP requirement.

• Journal of the Korean Society of Marine Environment & Safety
• /
• v.25 no.4
• /
• pp.468-475
• /
• 2019

An end-line flame arrester allows free venting in combination with flame protection for vertical vent applications. End-line flame arresters are employed in various fields, especially in shipping. In flame arresters, springs are essential parts because the spring load and the spring's elasticity determine the hood opening moment. In addition, the spring has to work under a high-temperature condition because of the burning gas flame. Therefore, it is necessary to analyze the mechanical load and elasticity of the spring when the flame starts to appear. Based on simulations of the working process of a specific end-line flame arrester, a thermal and structural analysis of the spring is performed. A three-dimensional model of a burned spring is built using computational fluid dynamics (CFD) simulation. Results of the CFD analysis are input into a finite element method simulation to analyze the spring structure. The research team focused on three cases of spring loads: 43, 93, and 56 kg, correspondingly, at 150 mm of spring deflection. Consequently, the spring load was reduced by 10 kg after 5 min under a $1,000^{\circ}C$ heat condition. The simulation results can be used to predict and estimate the spring's load and elasticity at the burning time variation. Moreover, the obtained outcome can provide the industry with references to optimize the design of the spring as well as that of the flame arrester.

• Journal of the Korean Society of Marine Environment & Safety
• /
• v.23 no.7
• /
• pp.918-925
• /
• 2017

Multiple reflections on exterior equipment with complex shape on naval ships cause unexpectedly high Radar Cross Section (RCS) distributions, and the directions of reradiated electromagnetic waves are hard to predict. Therefore, the optimum arrangement of exterior equipments should be considered according to the Radar Absorbing Structure (RAS) method. In this paper, the optimum arrangement for exterior equipments was determined to reduce multiple reflections and RCS even with complex shapes. The sequential descending arrangement method was used to establish an optimum arrangement algorithm. An LCS-2 type model was selected for optimum exterior equipment arrangements. In order to reduce computational cost, RCS distributions and multiple reflection path analysis of exterior equipments was carried out to select exterior equipments for optimum arrangement, and an optimum arrangement was determined to find positions with minimum RCS values. Also, the RCS reduction effect was analyzed using detectable radar range.

• Journal of the Korea Society of Computer and Information
• /
• v.26 no.4
• /
• pp.1-9
• /
• 2021

In this paper, we propose a training algorithm of support vector machine (SVM) with a sensitive variable. Although machine learning models enable automatic decision making in the real world applications, regulations prohibit sensitive information from being used to protect privacy. In particular, the privacy protection of the legally protected attributes such as race, gender, and disability is compulsory. We present an efficient least square SVM (LSSVM) training algorithm using a fully homomorphic encryption (FHE) to protect a partial sensitive attribute. Our framework posits that data owner has both non-sensitive attributes and a sensitive attribute while machine learning service provider (MLSP) can get non-sensitive attributes and an encrypted sensitive attribute. As a result, data owner can obtain the encrypted model parameters without exposing their sensitive information to MLSP. In the inference phase, both non-sensitive attributes and a sensitive attribute are encrypted, and all computations should be conducted on encrypted domain. Through the experiments on real data, we identify that our proposed method enables to implement privacy-preserving sensitive LSSVM with FHE that has comparable performance with the original LSSVM algorithm. In addition, we demonstrate that the efficient sensitive LSSVM with FHE significantly improves the computational cost with a small degradation of performance.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.12 no.6
• /
• pp.113-124
• /
• 2002

Proposals for a password-based authenticated key exchange protocol that have been published so far almost concentrated on the provable security. But in a real environment such as mobile one, efficiency is a critical issue as security. In this paper we discuss the efficiency of PAK which is secure in the random oracle model ［l］. Among 4 hash functions in PAK the instantiation for $H_1$, which outputs a verifier of the password, has most important effect on the computational efficiency. We analyze two different methods for $H_1$ suggested in ［1］ and we show that $H_{lq}$ has merits in transforming to EC or XTR variants as well as in the efficiency. As an efficient variant. we propose PAK2-EC and PAK2-XTR which do not require any additional step converting a hash output into a point of elliptic curve or XTR subgroup when compared to the previous work on the PAK［2］. Finally we compare PAK2 with the password-based authenticated key exchange protocols such as SPEKE, SRP, and AMP.

• Journal of Korea Water Resources Association
• /
• v.54 no.spc1
• /
• pp.1053-1060
• /
• 2021

Current water quality monitoring systems in Korea carried based on in-situ grab sample analysis. It is difficult to improve the current water quality monitoring system, i.e. shorter sampling period or increasing sampling points, because the current systems are both cost- and labor-intensive. One possible way to improve the current water quality monitoring system is to adopt a modeling approach. In this study, a modeling technique was introduced to support the current water quality monitoring system, and an artificial neural network model, the computational tool which mimics the biological processes of human brain, was applied to predict water quality of the river. The approach tried to predict concentrations of Total coliform at the outlet of the river and this showed, somewhat, poor estimations since concentrations of Total coliform were rapidly fluctuated. The approach, however, could forecast whether concentrations of Total coliform would exceed the water quality standard or not. As results, modeling approaches is expected to assist the current water quality monitoring system if the approach is applied to judge whether water quality factors could exceed the water quality standards or not and this would help proper water resource managements.

• Journal of Broadcast Engineering
• /
• v.26 no.4
• /
• pp.429-440
• /
• 2021

With the recent development of deep convolutional neural network learning, deep learning techniques applied to single image super-resolution are showing good results. One of the existing deep learning-based super-resolution techniques is RDN(Residual Dense Network), in which the initial feature information is transmitted to the last layer using residual dense blocks, and subsequent layers are restored using input information of previous layers. However, if all hierarchical features are connected and learned and a large number of residual dense blocks are stacked, despite good performance, a large number of parameters and huge computational load are needed, so it takes a lot of time to learn a network and a slow processing speed, and it is not applicable to a mobile system. In this paper, we use the residual dense structure, which is a continuous memory structure that reuses previous information, and the residual dense channel attention block using the channel attention method that determines the importance according to the feature map of the image. We propose a method that can increase the depth to obtain a large receptive field and maintain a concise model at the same time. As a result of the experiment, the proposed network obtained PSNR as low as 0.205dB on average at 4× magnification compared to RDN, but about 1.8 times faster processing speed, about 10 times less number of parameters and about 1.74 times less computation.

• Clean Technology
• /
• v.28 no.4
• /
• pp.316-322
• /
• 2022

The flow pattern at the inlet of the catalyst layer in a selective catalytic reduction (SCR) system is one of the key parameters influencing the performance of the denitrification process. In the curved diffusing parts between the ammonia injection grids and the catalyst layers, guide vanes are installed to improve flow uniformity. In the present study, a numerical simulation has been performed to investigate the effect of the geometrical configuration of the guide vanes on the aerodynamic characteristics of a denitrification facility. This application has been made to the existing SCR process in a large-scaled coal-fired power plant. The flow domain to be solved covers the whole region of the flow passages from the exit of the ammonia injection gun to the exit of the catalyst layers. ANSYS-Fluent was used to calculate the three-dimensional steady viscous flow fields with the proper turbulence model fitted to the flow characteristics. The root mean square of velocity and the pressure drop inside the flow passages were chosen as the key performance parameters. Four types of guides vanes were proposed to improve the flow quality compared to the current configuration. The numerical results showed that the type 4 configuration was the most effective at improving the aerodynamic performance in terms of flow uniformity and pressure loss.