• Proceedings of the Korean Information Science Society Conference
• /
• 2003.04d
• /
• pp.283-285
• /
• 2003

슈퍼컴퓨터 사용자들은 FTP(File Transfer Protocol)을 이용해서 대용량의 파일을 전송한다. 전송되는 데이터들은 승용차나 비행기 설계, 의약품 개발. 기상 예보 그리고 복잡한 수학적 계산 등과 같이 다양한 분야에 걸친 연구관련 데이터로서 고성능 슈퍼컴퓨터에 의한 연산 처리가 요구된다 기존의 FTP 는 네트워크 상태에 따라 전송 지연이나 데이터 손실 등의 문제로 사용자의 불편을 초래하였다. 이에 전송 성능을 효율적으로 높이고 데이터 손실을 최소화를 제공하는 연구가 필요하다. 근래의 TCP (Transmission Control Protocol) 성능 향상 연구에 관한 연구들의 관심은 크게 두 가지이다. 하나는 윈도우 사이즈 조절(auto-tuning)이고, 다른 하나는 Multi-stream 이다. 본 연구에서는 파일 전송 성능 향상을 위한 방법으로 윈도우 사이즈 조절 방법을 사용하였고, 네트워크 상태에 따라 QoS(quality of Service)를 제공한다. 이런 성능 향상 결과로 신뢰성 있는 네트워크를 제공하여 사용자들은 신속하게 데이터를 전송하며 연산처리 결과가 더욱 정확하다고 신뢰할 수 있다. 본 고에서는 대용량 파일을 전송 할 때 성능을 향상시키는 관련 연구를 알아보고 대용량 파일 전송 중 네트워크 상태에 따라 005를 능동적으로 작용하여 테스트하고 성능을 분석하였다.

• Journal of the Korean Society of Manufacturing Process Engineers
• /
• v.21 no.7
• /
• pp.43-52
• /
• 2022

Design of a controller for a high-precision servo control system has been a popular topic while finding optimal parameters for multiple controllers is still a challenging subject. In this paper, we propose a practical scheme to optimize multi-parameters for the robust servo controller design by introducing a new cost function and optimization scheme. The proposed design method provides a simple and practical tool for the systematic servo design to reduce the control error with guaranteeing robust stability of the overall system. The reduction of the position error by 24% along with a faster convergence rate is demonstrated using a typical hard disk drive servo controller with 41 parameters.

• Annual Conference on Human and Language Technology
• /
• 2020.10a
• /
• pp.172-175
• /
• 2020

최근 자연어처리는 대용량 코퍼스를 이용하여 언어 모델을 사전 학습하고 fine-tuning을 적용함으로 다양한 태스크에서 최고 성능을 갱신하고 있다. BERT기반의 언어 모델들은 양방향의 Transformer만 모델링 되어 있지만 BART는 양방향의 Transformer와 Auto-Regressive Transformer가 결합되어 사전학습을 진행하는 모델로 본 논문에서는 540MB의 코퍼스를 이용해 한국어 BART 모델을 학습 시키고 여러 한국어 자연어처리 태스크에 적용하여 성능 향상 있음을 보였다.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.10 no.4
• /
• pp.659-666
• /
• 2006

Servo-motors are used as key components of automated system by performing precise motion control as accurate positioning and accurate speed regulation in response to the commands from computers and sensors. Especially, the linear brushless servo-motors have numerous advantages over the rotary servo motors which have connection with the friction induced transfer mechanism such as ball screws, timing belts, rack/pinion. This paper proposes an estimation method of unknown motor system parameters using the informations from the sinusoidal driving type linear brushless DC motor dynamics and outputs. The estimated parameters can be used to tune the controller gain and a disturbance observer. In order to meet this purpose high performance Digital Signal Processor, TMS320F240, designed originally for implementation of a Field Oriented Control(FOC) technology is adopted as a controller of the liner BLDC servo motor. Having A/D converters, PWM generators, rich I/O port internally, this servo motor application specific DSP play an important role in servo motor controller. This linear BLDC servo motor system also contains IPM(Intelligent Power Module) driver and hail sensor type current sensor module, photocoupler module for isolation of gate signals and fault signals.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.12 no.4
• /
• pp.353-360
• /
• 2002

Methods for automatic tuning of PID controllers have been on of the results of the active research on control. The proposed controller also is auto-tuning of PID controller The proposed immune algorithm has an uncomplicated structure and memory-cell mechanism as the optimization algorithm which imitates the principle of humoral immune response. We use the proposed algorithm to solve optimization of PID controller parameters. Up to now, the applications of immune algorithm have been optimization problems with non-varying system parameters. Therefore the usefulness of memory-cell mechanism in immune algorithm is without. And research of memory-cell mechanism does not give us entire satisfaction. This paper proposes the immune algorithm using a memory-cell mechanism which can be the application of system with nonlinear varying parameters. To verify performance of the proposed immune algorithm, the speed control of nonlinear DC motor are performed. The results of Computer simulations represent that the proposed immune algorithm shows a fast convergence speed and a good control performances under the varying system parameters.

• Journal of Power Electronics
• /
• v.3 no.2
• /
• pp.99-114
• /
• 2003

This paper presents a feasible development on a highly accurate quick response adjustable speed drive implementation fur general purpose induction motor which operates on the basis of sensorless slip frequency type vector controlled sine-wave PWM inverter with an automatic tuning machine parameter estimation schemes. In the first place, the sensorless vector control theory on the three-phase voltage source-fed inverter induction motor drive system is developed in slip frequency based vector control principle. In particular, the essential procedure and considerations to measure and estimate the exact stator and rotor circuit parameters of general purpose induction motor are discussed under its operating conditions. The speed regulation characteristics of induction motor operated by the three-phase voltage-fed type current controlled PWM inverter using IGBT's is illustrated and evaluated fur machine parameter variations under the actual conditions of low frequency and high frequency operations for the load torque. In the second place, the variable speed induction motor drive system, employing sensorless vector control scheme which is based on three -phase high frequency carrier PWM inverter with automatic toning estimation schemes of the temperature -dependent and -independent machine circuit parameters, is practically implemented using DSP-based controller. Finally, the dynamic speed response performances for largely changed load torque disturbances as well as steady state speed vs. torque characteristics of this induction motor control implementation are illustrated and discussed from an experimental point of view.

• Journal of the Korean Society of Radiology
• /
• v.18 no.3
• /
• pp.267-273
• /
• 2024

Using a cloud-based vertex AI platform that can develop an artificial intelligence learning model without coding, this study easily developed an artificial intelligence learning model by the non-professional general public and confirmed its clinical applicability. Nine dental diseases and 2,999 root disease X-ray images released on the Kaggle site were used for the learning data, and learning, verification, and test data images were randomly classified. Image classification and multi-label learning were performed through hyper-parameter tuning work using a learning pipeline in vertex AI's basic learning model workflow. As a result of performing AutoML(Automated Machine Learning), AUC(Area Under Curve) was found to be 0.967, precision was 95.6%, and reproduction rate was 95.2%. It was confirmed that the learned artificial intelligence model was sufficient for clinical diagnosis.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2004.10a
• /
• pp.818-821
• /
• 2004

With a rapid development in the area of micro and ultra precision technology, the micro surface machining of small size parts are explosively increased. Especially, to improve efficiency of various beams in lens and reflector, non-rotational symmetric form and several mm level heights changeable surface can be machined at a time. These geometric complex 3D surface cannot be machined by general short stroke FTS. The long stroke FTS if firmly needed to move directly several mm and have nm level positioning accuracy for the complex surface form. The long stroke FTS used linear motors to drive moving unit long and fine, aero static bearings to decrease friction and moving errors in guide way, optical linear scale with nm level resolution to measure position of FTS. Furthermore, to increase the performance of acceleration of FTS, the light material, such as AL is used for the structure and the high stiffness box type structure is selected. In this paper, the genetic algorithm approach is described to determine a set of design parameters for auto tuning. The authors have attempted to model the design problem with the objective of minimizing the error, such as variable pattern change. This method can give the better alternative than existing other method.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.8 no.8
• /
• pp.1195-1200
• /
• 2013

In this paper, we proposed an enhanced Max-Min neural network by auto-tuning of learning rate using fuzzy control method. For the reduction of training time required in the competition stage, the method was proposed that arbitrates dynamically the learning rate by applying the numbers of the accuracy and the inaccuracy to the input of the fuzzy control system. The experiments using real concrete crack images showed that the enhanced Max-Min neural network was effective in the recognition of direction of the extracted cracks.

• Progress in Superconductivity
• /
• v.5 no.1
• /
• pp.55-60
• /
• 2003

We designed and constructed integrated 3-channel flux-locked-loop (FLL) electronic system for the control and readout of high-T$_{c}$ SQUIDs. This system consists of low noise preamplifiers, integrators, interface circuits, and software. FLL operation was carried out with biased signals of 19 KHz modulated current and 150 KHz modulated flux, which are reconstructed as detected signals by preamplifier and demodulator. Computer controlled interface circuits regulate FLL circuit and adjust SQUID parameters to the optimum operating condition. The software regulates interface circuits to make an auto-tuning for the control of SQUIDs, and displays readout data from FLL circuit. 3-channel SQUID electronic system was assembled with 3 FLL-interface circuit boards and a power supply board in the aluminum case of 56 mm ${\times}$ 53 mm${\times}$ 150 mm. Overall noise of the system was around 150 fT/(equation omitted)Hz when measured in the shielded room, 200 fT/(equation omitted)Hz in a weakly shielded room, respectively.y.