• Smart Media Journal
• /
• v.9 no.4
• /
• pp.36-43
• /
• 2020

The Monte Carlo tree search (MCTS) is a popular method for implementing an intelligent game program. It has several hyper-parameters that require an optimization for showing the best performance. Due to the stochastic nature of the MCTS, the hyper-parameter optimization is difficult to solve. This paper uses the self-playing capability of the MCTS-based game program for optimizing the hyper-parameters. It seeks a winner path over the hyper-parameter space while performing the self-play. The top-q longest winners in the winner path compete for the final winner. The experiment using the 15-15-5 game (Omok in Korean name) showed a promising result.

• Journal of KIISE:Software and Applications
• /
• v.29 no.11
• /
• pp.785-794
• /
• 2002

Automatic stock trading systems should be able to solve various kinds of optimization problems such as market trend prediction, stock selection, and trading strategies, in a unified framework. But most of the previous trading systems based on supervised learning have a limit in the ultimate performance, because they are not mainly concerned in the integration of those subproblems. This paper proposes a stock trading system, called R-Trader, based on reinforcement teaming, regarding the process of stock price changes as Markov decision process (MDP). Reinforcement learning is suitable for Joint optimization of predictions and trading strategies. R-Trader adopts two popular reinforcement learning algorithms, temporal-difference (TD) and Q, for selecting stocks and optimizing other trading parameters respectively. Technical analysis is also adopted to devise the input features of the system and value functions are approximated by feedforward neural networks. Experimental results on the Korea stock market show that the proposed system outperforms the market average and also a simple trading system trained by supervised learning both in profit and risk management.

• Korean Journal of Microbiology
• /
• v.46 no.1
• /
• pp.46-51
• /
• 2010

Coenzyme Q10 (CoQ10) is an essential lipid-soluble component of membrane-bound electron transport chains. CoQ10 is involved in several aspects of cellular metabolism and is increasingly being used in therapeutic applications for several diseases. Despite the recent accomplishments in metabolic engineering of Escherichia coli for CoQ10 production, the production levels are not yet competitive with those by fermentation or isolation. So we tested several microorganisms obtained from the KCTC of Biological Resource Center to find novel sources of strain-development for CoQ10-production. Then we selected two strains, Paracoccus denitrificans (KCTC 2530) and Asaia siamensis (KCTC 12914), and tested to optimize the CoQ10 production conditions. Among the carbon sources tested, CoQ10 production was the highest when fructose was supplied about 4% concentration. Yeast extract produced the highest CoQ10 production about 2% concentration. The highest CoQ10 production was obtained at pH 6.0 for P. denitrificans and pH 8.0 for A. siamensis. And two strains showed the highest CoQ10 production at $30^{\circ}C$, but the highest DCW was obtained at $37^{\circ}C$. In the fed-batch culture, P. denitrificans yielded $14.34{\pm}0.473$ mg and A. siamensis yielded $12.53{\pm}0.231$ mg of final CoQ10 production.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.11 no.4
• /
• pp.23-28
• /
• 1974

One of the consistent nuisances in accurate design of circuits containing coils with core is how to take into account the frequency dependence of Q of actual coils. The conventional equivalent circuit consisting of an inductance and a series (constant) resistance and possibly a parallel (constant) capacitance is of little use in this situation since the core loss itself is strongly dependent on the frequency. In order to circumvent this difficulty, in this paper, a mathematical expression for Q of a given core as a function of inductance and frequency is first assumed and parameters in this expression are optimiged so as to best fit the data provided by the core manufacturer or obtained experimentally. This expression is then utilized in accurate calculation of the frequency response of a given circuit required in the optimal design of circuits containing coils. In other words the proposed approach is an effective combination of an approximate expression of coil's Q and circuit optimisation technique, which seems to have solved, to a great extent, the stated difficulty associated with actual coils. As for the optimization technique, ths Fletcher-Powell procedure was employed and one example was given to illustrate the proposed approach.

• Journal of the Korean Institute of Illuminating and Electrical Installation Engineers
• /
• v.21 no.5
• /
• pp.1-9
• /
• 2007

This paper is proposed an efficiency optimization control algorithm for a synchronous reluctance motor which minimizes the coner and iron losses. The design of the speed controller based on adaptive fuzzy-neural networks(AFNN) controller that is implemented using fuzzy control and neural networks. There exists a variety of combinations of d and q-axis current which provide a specific motor torque. The objective of the efficiency optimization controller is to seek a combination of d and q-axis current components, which provides minimum losses at a certain operating point in steady state. The proposed algorithm allows the electromagnetic losses in variable speed and torque drives to be reduced while keeping good torque control dynamics. The control performance of the hybrid artificial intelligent controller is evaluated by analysis for various operating conditions. Analysis results are presented to show the validity of the proposed algorithm.

• Proceedings of the KIEE Conference
• /
• 2009.04b
• /
• pp.29-31
• /
• 2009

Optimal efficiency control of synchronous reluctance motor(SynRM) is very important in the sense of energy saving and conservation of natural environment because the efficiency of the SynRM is generally lower than that of other types of AC motors. This paper is proposed an efficiency optimization control for the SynRM which minimizes the copper and iron losses. The design of the speed controller based on fuzzy-neural networks (FN)-PI controller that is implemented using fuzzy control and neural networks. There exists a variety of combinations of d and q-axis current which provide a specific motor torque. The objective of the efficiency optimization control is to seek a combination of d and q-axis current components, which provides minimum losses at a certain operating point in steady state. It is shown that the current components which directly govern the torque production have been very well regulated by the efficiency optimization control scheme. The proposed algorithm allows the electromagnetic losses In variable speed and torque drives to be reduced while keeping good torque control dynamics. The control performance of the proposed controller is evaluated by analysis for various operating conditions. Analysis results are presented to show the validity of the proposed algorithm.

• Journal of Power System Engineering
• /
• v.21 no.6
• /
• pp.56-61
• /
• 2017

The check valves are mechanical valves that permit fluids to flow in only one direction, preventing flow from reversing. It is classified as one way directional valves. There are various types of check valves that used in a marine application. A lift type check valve uses the disc to open and close the passage of fluid. The disc lift up from seat as pressure below the disc increases, while drop in pressure on the inlet side or a build up of pressure on the outlet side causes the valve to close. An important concept in check valves is the cracking pressure which is the minimum upstream pressure at which the valve will operate. On the other hand, optimization is a process of finding the best set of parameters to reach a goal while not violating certain constraints. The AMESim software provides NLPQL(Nonlinear Programming by Quadratic Lagrangian) and genetic algorithm(GA) for optimization. NLPQL is the implementation of a SQP(sequential quadratic programming) algorithm. SQP is a standard method, based on the use of a gradient of objective functions and constraints to solve a non-linear optimization problem. A characteristic of the NLPQL is that it stops as soon as it finds a local minimum. Thus, the simulation results may be highly dependent on the starting point which user give to the algorithm. In this paper, we carried out optimization design of the check valve with NLPQL algorithm.

• Journal of Communications and Networks
• /
• v.13 no.4
• /
• pp.327-338
• /
• 2011

With the incitation to reduce power consumption and the aggressive reuse of spectral resources, there is an inevitable trend towards the deployment of small-cell networks by decomposing a traditional single-tier network into a multi-tier network with very high throughput per network area. However, this cell size reduction increases the complexity of network operation and the severity of cross-tier interference. In this paper, we consider a downlink two-tier network comprising of a multiple-antenna macrocell base station and a single femtocell access point, each serving multiples users with a single antenna. In this scenario, we treat the following beamforming optimization problems: i) Total transmit power minimization problem; ii) mean-square error balancing problem; and iii) interference power minimization problem. In the presence of perfect channel state information (CSI), we formulate the optimization algorithms in a centralized manner and determine the optimal beamformers using standard convex optimization techniques. In addition, we propose semi-decentralized algorithms to overcome the drawback of centralized design by introducing the signal-to-leakage plus noise ratio criteria. Taking into account imperfect CSI for both centralized and semi-decentralized approaches, we also propose robust algorithms tailored by the worst-case design to mitigate the effect of channel uncertainty. Finally, numerical results are presented to validate our proposed algorithms.

• Journal of Auto-vehicle Safety Association
• /
• v.7 no.3
• /
• pp.7-13
• /
• 2015

Occupant protection performance in frontal crashes has been developed and assessed for mainly front seat occupants over many years, and in recent years protection of rear seat occupants has also been extensively discussed. Unlike the front seats, the rear seats are often occupied by children seated in rear-facing or forward - facing child restraint systems, or booster seats. In the ENCAP, child occupant protection assessments using 18-month-old(P1.5) and 3-year-old(P3) test dummies in the rear seat have already been changed to new type of 18-month-old (Q1.5)and 3-year-old(Q3) test dummies. In addition, ENCAP are scheduled with the development and introduction of test dummies of 6-year-old (Q6) and 10.5-year-old children(Q10) starting 2016. In KNCAP, Q6 and Q10 child dummies will be introduced in 2017 as well. Automobile manufacturers need to develop safety performance for new child dummies closely. In this paper, we focused on Q6 and Q10 child dummies sitting in child restraint system. Offset frontal crash tests were conducted using two types of test dummies, Q6 and Q10 child dummies, positioned in the rear seat. Q6 and Q10 were used to compare dummy kinematics in rear seating positions between Q6 behind the driver's seat and Q10 behind the front passenger's seat. The full vehicle sled test results of both dummies were conducted with different restraint systems. It showed that several injury and image data was collected as the result of the full vehicle sled test. Based on the results of these investigations, this paper describes which factor is most important and combination is the best performance when evaluating rear seat occupant protection for Q6 and Q10 child dummies.

• 제어로봇시스템학회:학술대회논문집
• /
• 2003.10a
• /
• pp.1300-1304
• /
• 2003

The following paper deals with a control problem of a mobile robot in home network environment. The home network causes the mobile robot to communicate with sensors to get the sensor measurements and to be adapted to the environment changes. To get the improved performance of control of a mobile robot in spite of the change in home network environment, we use the fuzzy inference system with multiple reward reinforcement learning. The multiple reward reinforcement learning enables the mobile robot to consider the multiple control objectives and adapt itself to the change in home network environment. Multiple reward fuzzy Q-learning method is proposed for the multiple reward reinforcement learning. Multiple Q-values are considered and max-min optimization is applied to get the improved fuzzy rule. To show the effectiveness of the proposed method, some simulation results are given, which are performed in home network environment, i.e., LAN, wireless LAN, etc.