• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.16 no.10 s.101
• /
• pp.965-972
• /
• 2005

In this paper, we first review LDPC codes in general and a belief propagation algorithm that works in logarithm domain. LDPC codes, which is chosen for second generation digital video broadcasting standard, are required a large number of computation due to large size of coded block and iteration. Therefore, we presented two kinds of low computational algorithm for LDPC codes. First, sequential decoding with partial group is proposed. It has same H/W complexity, and fewer number of iteration's are required at same performance in comparison with conventional decoder algerian. Secondly, early detection method for reducing the computational complexity is proposed. Using a confidence criterion, some bit nodes and check node edges are detected early on during decoding. Through the simulation, we knew that the iteration number are reduced by half using subset algorithm and computational complexity of early detected method is about $50\%$ offs in case of check node update, $99\%$ offs in case of check node update compared to conventional scheme.

• Journal of Intelligence and Information Systems
• /
• v.5 no.1
• /
• pp.95-101
• /
• 1999

In this paper, a comparison between different methods to combine predictions from neural networks will be given. These methods are bagging, bumping, and balancing. Those are based on the analysis of the ensemble generalization error into an ambiguity term and a term incorporating generalization performances of individual networks. Neural Networks and AI machine learning models are prone to overfitting. A strategy to prevent a neural network from overfitting, is to stop training in early stage of the learning process. The complete data set is spilt up into a training set and a validation set. Training is stopped when the error on the validation set starts increasing. The stability of the networks is highly dependent on the division in training and validation set, and also on the random initial weights and the chosen minimization procedure. This causes early stopped networks to be rather unstable: a small change in the data or different initial conditions can produce large changes in the prediction. Therefore, it is advisable to apply the same procedure several times starting from different initial weights. This technique is often referred to as training ensembles of neural networks. In this paper, we presented a comparison of three statistical methods to prevent overfitting of neural network.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.30 no.6C
• /
• pp.442-449
• /
• 2005

In this paper, we introduce three kinds of simplified high-speed decoding algorithms for turbo product decoder. First, A parallel decoder structure, the row and column decoders operate in parallel, is proposed. Second, HAD(Hard Decision Aided) algorithm is used for early-stopping algorithm. Lastly, P-Parallel TPC decoder is a parallel decoding scheme, processing P rows and P columns in parallel instead of decoding one by one as that in the original scheme.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.32 no.4C
• /
• pp.379-388
• /
• 2007

In this paper, we propose an adaptive turbo decoding algorithm for high order modulation scheme combined with originally design for a standard rate-1/2 turbo decoder for B/QPSK modulation. A transformation applied to the incoming I-channel and Q-channel symbols allows the use of an off-the-shelf B/QPSK turbo decoder without any modifications. Adaptive turbo decoder process the received symbols recursively to improve the performance. As the number of iterations increase, the execution time and power consumption also increase as well. The source of the latency and power consumption reduction is from the combination of the radix-4, dual-path processing, parallel decoding, and early-stop algorithms. We implemented the proposed scheme on a field-programmable gate array (FPGA) and compared its decoding speed with that of a conventional decoder. From the result of implementation, we confirm that the decoding speed of proposed adaptive decoding is faster than conventional scheme by 6.4 times.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 1997.04a
• /
• pp.128-132
• /
• 1997

The conventional fire detection devices are operated after a processed fire phase, which are sensing only a high density of smoke level or high temperature heat. They are not so precision to detect a fire in the early phase to protect the facility from the fire. We need to develope a new high precision smoke detection system to keep expensive industial facilities most reliably from fire. A new optical precision smoke detection system was developed. It monitors very low level density of smoke psrticles in the air. It is operated continuously through many years without a stop or any malfunction. The developed precision smoke detection system will be installed in important industrial facilities,such as power plants, underground common tunnel,main control rooms,computer rooms etc.

• Journal of the Korean Home Economics Association
• /
• v.30 no.2
• /
• pp.35-43
• /
• 1992

This study is to investigate the problems of pollution caused by drycleaning solvents. From the problem of each solvents, safety count-step have to set up. It is following. 1) Solvent: Smong phenomenon of optical chemistry. It has to develop safety tumbler that is attached retrievable device. 2) Perchloroethylene: Pollution of underground water and cause of cancer. Considering the influence on environment and human bing, it has to set up more strong safty plan than now. 3) Chlorofluorocabon : Destory of ozon layer and green house effect on earth. It is being to stop the production until the early part of the year 2000. So it is developing a substitute for chlorofluorocarbon. 4) 1,1,1-trichloroethane : Pollution of underground water and destroy of ozon layzer. At the end of the year 2004, there will be end production of it.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.30 no.6C
• /
• pp.450-456
• /
• 2005

In this paper, we propose a high-speed turbo decoding algorithm and present results of its implementation. The latency caused by (de)interleaving and iterative decoding in conventional MAP turbo decoder can be dramatically reduced with the proposed scheme. The main cause of the time reduction is to use radix-4, center to top, and parallel decoding algorithm. The reduced latency makes it possible to use turbo decoder as a FEC scheme in the real-time wireless communication services. However the proposed scheme costs slight degradation in BER performance because the effective interleaver size in radix-4 is reduced to an half of that in conventional method. To ensure the time reduction, we implemented the proposed scheme on a FPGA chip and compared with conventional one in terms of decoding speed. The decoding speed of the proposed scheme is faster than conventional one at least by 5 times for a single iteration of turbo decoding.

• Proceedings of the KOSOMBE Conference
• /
• v.1996 no.11
• /
• pp.353-356
• /
• 1996

Early this year the construction of new high voltage power lines have been forced to stop due to the demonstration of residents in Kwa-chun. According to mutual agreement between the residents and Korea Electric Power Corporation, electromagnetic field(EMF) from present power lines was measured and EMF after completion of new power lines was estimated. The results show that EMF near power lines Is relatively low, and EMFnear electric substation is quite high due to the short distance between the power lines and ground. For 4 conduction 345kV line, magnetic field decreases to 3mG at the distance of 50m from the power line.

• Journal of Chest Surgery
• /
• v.18 no.4
• /
• pp.812-816
• /
• 1985

In the past several years the popularity of the motor cycle has produced an increasing incidence of the injuries to the larynx and trachea. Most of all on accidents come to death and survivors to the hospital are rare. Early diagnosis and to keep air way are necessary to initiate proper treatment in injury of upper air way. Meticulous apposition of mucous membrane and reconstitution of laryngeal skeleton are important. We experienced a rare case of 26 year old men with cricothyroidal transection after trauma. On Oct. 17, 1985, the patient struck his neck on baggage frame of truck when dropping from his motor cycle on sudden stop. Emergency tracheal intubation on distal segment of trachea was accomplished by otolaryngologist in a local clinic. He was transferred to our hospital. Exploration 2 hours later revealed complete separation of cricoid cartilage from thyroid cartilage. The recurrent laryngeal nerve could not be identified. Anastomosis of thyroid and cricoid was accomplished and Portex endotracheal tube was inserted as splint for 10 days. No stenosis developed. The air way appeared adequate for moderate physical activity though paramedian fixation of vocal cord paralysis. Postoperative follow-up course has been good after he discharged on POD 14 days.

• ETRI Journal
• /
• v.43 no.6
• /
• pp.1004-1012
• /
• 2021

The collaboration productively interacting between multi-agents has become an emerging issue in real-world applications. In reinforcement learning, multi-agent environments present challenges beyond tractable issues in single-agent settings. This collaborative environment has the following highly complex attributes: sparse rewards for task completion, limited communications between each other, and only partial observations. In particular, adjustments in an agent's action policy result in a nonstationary environment from the other agent's perspective, which causes high variance in the learned policies and prevents the direct use of reinforcement learning approaches. Unexpected social loafing caused by high dispersion makes it difficult for all agents to succeed in collaborative tasks. Therefore, we address a paradox caused by the social loafing to significantly reduce total returns after a certain timestep of multi-agent reinforcement learning. We further demonstrate that the collaborative paradox in multi-agent environments can be avoided by our proposed effective early stop method leveraging a metric for social loafing.