• Journal of the Institute of Electronics and Information Engineers
• /
• v.53 no.4
• /
• pp.79-86
• /
• 2016

Physical layer security is cryptography technique to protect information by using physical nature of signals. Currently, many works on physical layer security have been actively researching while those researching models still have some problems to be solved. Eavesdropper does not share its channel state information with legitimate users to hide its presence. And when node transmits signal, hardware impairments are occurred, whereas many current researches assume that node model is ideal node and does not consider hardware impairments. The main features and contributions of this paper to solve these problems are as follows. First, our proposed system model deploys torch node around legitimate user to obtain channel state information of eavesdropper and considers hardware impairments by using channel state information of torch node. Second, we derive closed-form expression of intercept probability for the proposed system model. The results of the performance evaluation through various simulations to find out the effects on proposed system model in physical layer security show that imperfect channel state information does not effect on intercept probability while imperfect node model effects on intercept probability, Ergodic secrecy capacity and secrecy capacity.

• 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 the Computational Structural Engineering Institute of Korea
• /
• v.33 no.6
• /
• pp.359-366
• /
• 2020

In this study, a simulation-based damage estimation method for helidecks is proposed using an artificial neural network. The structural members that share a connecting node in the helideck are regarded as a damage group, and a total of 37,400 damage scenarios are numerically generated by applying randomly assigned damage to up to three damage groups. Modal analysis is then performed for all the damage scenarios, which are selectively used as either training or validation or verification sets based on the purpose of use. An artificial neural network with three hidden layers is constructed using a PyTorch program to recognize the patterns of the modal responses of the helideck model under both damaged and undamaged states, and the network is successively trained to minimize the loss function. Finally, the estimated damage rate from the proposed artificial neural network is compared to the actual assigned damage rate using 400 verification scenarios to show that the neural network is able to estimate the location and amount of structural damage precisely.