• The KIPS Transactions:PartB
• /
• v.11B no.2
• /
• pp.207-212
• /
• 2004

This paper presents a reinforcement learning framework for stock trading systems. Trading system parameters are optimized by Q-learning algorithm and neural networks are adopted for value approximation. In this framework, cooperative multiple agents are used to efficiently integrate global trend prediction and local trading strategy for obtaining better trading performance. Agents Communicate With Others Sharing training episodes and learned policies, while keeping the overall scheme of conventional Q-learning. Experimental results on KOSPI 200 show that a trading system based on the proposed framework outperforms the market average and makes appreciable profits. Furthermore, in view of risk management, the system is superior to a system trained by supervised learning.

• Korean Journal of Mineralogy and Petrology
• /
• v.34 no.3
• /
• pp.157-167
• /
• 2021

Lead (Pb) is one of the key trace elements, exhibiting a peculiar partitioning behavior into silicate melts in contact with minerals. Partitioning behaviors of Pb between silicate mineral and melt have been known to depend on melt composition and thus, the atomic structures of corresponding silicate liquids. Despite the importance, detailed structural studies of Pb-bearing silicate melts are still lacking due to experimental difficulties. Here, we explored the effect of lead content on the atomic structures, particularly the evolution of silicate networks in Pb-bearing sodium metasilicate ([(PbO)_x(Na₂O)_1-x]·SiO₂) glasses as a model system for trace metal bearing natural silicate melts, using ²⁹Si solid-state nuclear magnetic resonance (NMR) spectroscopy. As the PbO content increases, the ²⁹Si peak widths increase, and the maximum peak positions shift from -76.2, -77.8, -80.3, -81.5, -84.6, to -87.7 ppm with increasing PbO contents of 0, 0.25, 0.5, 0.67, 0.86, and 1, respectively. The ²⁹Si MAS NMR spectra for the glasses were simulated with Gaussian functions for Qⁿ species (SiO₄ tetrahedra with n BOs) for providing quantitative resolution. The simulation results reveal the evolution of each Qⁿ species with varying PbO content. Na-endmember Na₂SiO₃ glass consists of predominant Q² species together with equal proportions of Q¹ and Q³. As Pb replaces Na, the fraction of Q² species tends to decrease, while those for Q¹ and Q³ species increase indicating an increase in disproportionation among Qⁿ species. Simulation results on the ²⁹Si NMR spectrum showed increases in structural disorder and chemical disorder as evidenced by an increase in disproportionation factor with an increase in average cation field strengths of the network modifying cations. Changes in the topological and configurational disorder of the model silicate melt by Pb imply an intrinsic origin of macroscopic properties such as element partitioning behavior.

• Journal of the Mineralogical Society of Korea
• /
• v.21 no.3
• /
• pp.321-329
• /
• 2008

Amorphous silica nanoparticles are among the most fundamental $SiO_2$ compounds, having implications in diverse geological processes and technological applications. Here, we explore structural details of amorphous silica nanoparticles with varying particle sizes (7 and 14 nm) using $^{29}Si$ and $^{1}H$ MAS NMR spectroscopy together with quantum chemical calculations to have better prospect for their size-dependent atomic structures. $^{29}Si$ MAS NMR spectra at 9.4 T resolve $Q^2,\;Q^3$ and $Q^4$ species at -93 ppm, -101 ppm, -110 ppm, respectively. The fractions of $Q^2,\;Q^3,\;O^4$ species are $7{\pm}1%,\;27{\pm}2%$, and $66{\pm}2%$ for 7 nm amorphous silica nanoparticles and $6{\pm}1%,\;21{\pm}2%$, and $73{\pm}2%$ for 14 nm amorphous silica nanoparticles. Whereas it has been suggested that $Q^2$ and $Q^3$ species exist on particles surfaces, the difference in $Q^{2}\;+\;Q^{3}$ fraction in both 7 and 14 nm particles is not significant, suggesting that $Q^2$ and $Q^3$ species could exist inside particles. $^{1}H$ MAS NMR spectra at 11.7 T shows diverse hydrogen environments, including physisorbed water, hydrogen bonded silanol, and non-hydrogen bonded silanol with varying hydrogen bond strength. The hydrogen contents in the 7nm silica nanoparticles (including water and hydroxyl groups) are about 3 times of that of 14 nm particles. The larger chemical shills for proton environments in the former suggest stronger hydrogen bond strength. The fractions of non-hydrogen bonded silanols in the 14 nm amorphous silica nanoparticles are larger than those in 7 nm amorphous silica nanoparticles. This observation suggests closer proximity among hydrogen atoms in the nanoparticles with smaller diameter. The current results with high-resolution solid-state NMR reveal previously unknown structural details in amorphous silica nanoparticles with particle size.

• Journal of the Institute of Electronics Engineers of Korea TC
• /
• v.47 no.2
• /
• pp.62-66
• /
• 2010

In this article, a oscillator at X-band with a double H-shape metamaterial resonator (DHMR) based on high-Q is proposed with metamaterial structure to improve Ihe phase noise and output power. The proposed oscillator is required low phase noise and high output power for the high performance frequency synthesizer. DHMR is designed to be high-Q at resonance frequency through strong coupling of E-field. This character makes phase noise excellent. The oscillator using DHMR is oscillated in X-band so as to apply frequency synthesizer of radar systems. The output power is 4.33 dBm and the phase noise is -108 dBc/Hz at 100 kHz offset of carrier frequency.

• Journal of Korean Society of Forest Science
• /
• v.83 no.2
• /
• pp.155-163
• /
• 1994

This study was conduced to provide a method of tending practices for natural mixed stands, which have a possibility of inducing to selection cutting stand based on the analysis of stand structures. For this, first, diameter distribution for a balanced selection cutting stand was predicted using Q theory applied to natural stands. The method of estimating diminution ratio coefficient Q is presented. The possible tending practice method which could be adopted to natural stands is discussed based on two different diameter distributions and side conditions.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2003.05a
• /
• pp.85-88
• /
• 2003

강화학습의 한가지 방법인 Q-learning은 최근에 Linear Quadratic Regulation(이하 LQR) 문제에 성공적으로 적용된 바 있다. 특히, 시스템 모델의 파라미터에 대한 구체적인 정보없이 적절한 입ㆍ출력만으로 학습을 통해 문제의 해결이 가능하므로 상황에 따라 매우 실용적인 방법이 될 수 있다. 뉴럴-큐 기법은 이러한 Q-learning의 Q-value를 MLP(multilayer perceptron) 신경망의 출력으로 대치시켜, 비선형 시스템의 최적제어 문제를 다룰 수 있게 한 방법이다. 그러나, 뉴럴-큐 기법은 신경망의 구조를 먼저 결정한 후 역전파 알고리즘을 이용해 학습하는 절차를 행하므로, 시행착오를 통해 신경망 구조를 결정해야 한다는 점, 역전파 알고리즘의 적용에 따라 신경망의 연결강도 값들이 지역적 최적해로 수렴한다는 점등의 문제점이 있다. 본 논문에서는 뉴럴-큐 학습의 도구로 KFD회귀를 이용하여 Q 함수의 근사 기법을 제안하고 관련 수식을 유도하였다. 그리고, 모의 실험을 통하여, 제안된 뉴럴-큐 방법의 적용 가능성을 알아보았다.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.17 no.1
• /
• pp.121-125
• /
• 2007

Differential cryptanalysis and linear cryptanalysis are the most powerful approaches known for attacking many block ciphers and used to evaluating the security of many block ciphers. So designers have designed secure block ciphers against these cryptanalyses. In this paper, we present new three block cipher structures. And for given r, we prove that differential(linear) probabilities for r-round blockcipher structures are upper bounded by $p^2(q^2),\;2p^2(2q^2)$ if the maximum differential(linear) probability is p(q) and the round function is a bijective function.

• Korean Journal of Environment and Ecology
• /
• v.8 no.2
• /
• pp.93-106
• /
• 1995

To investigate the forest structure in Ijontong Cholto area, twenty four plots were set up (each l0$\times$10m ) and surveyed. According to the analysis of ordination by DCA, community divided into 5 groups. The divided groups are Quercus serrata -Q. valiabilis community(I), Q. serrata - Pinus densiflora community(II), Q. mongolica - Lindera obtusiloba community(III), Carpinus laxiflora - Acer pseudo-sieboldianum community(IV) and Acer mono community(V). So the successional trends seems to be from P. densiflora through Q. serrata, Q. valiabilis, Q. mongolica to C. laxiflora, C. cordata, A. mono in canopy layer. And in the subtree and shrub layer, it was expected that from Rhododendron mucronulatum, Rh. schlippenbachii to A. pseudo-sieboldianum, L. obtusiloba. But some sites were likely to become Q. mongolica edaphic climax.

• Korean Journal of Environment and Ecology
• /
• v.19 no.2
• /
• pp.90-98
• /
• 2005

To investigate the vegetation structure of the Woraksan National Park, twenty-two plots were set up and surveyed. According to the analysis of classification by TWINSP AN, the communities were divided by six groups; Pinus densiflora-Quercus serrata, Q. mongolica-Q. serrata, Q. mongolica-Betula davueica, P. densiflora-Q. mongolica, Q. mongolica, Q. mongolica-Acer pseudosieboldianum. Successional sere of the forest vegetation in the surveyed area were proeeding from Pinus densiflora to Quercus serrata, Quercus monogolica in the canopy layer group and from Rhododendron Rhododendron schlippenbachii to Fraxinus sieboldiana and Acer pseudosieboldianum in the understory layer group. But Q. mongolica might be edaphic climax species in some area.

• Transactions of the Korean Society of Mechanical Engineers B
• /
• v.36 no.8
• /
• pp.829-837
• /
• 2012

Direct numerical simulations were carried out for turbulent channel flows with $Re_{\tau}$ = 180, 395 and 590 to investigate the turbulent flow structure related to the Reynolds shear stress. By examining the probability density function, the second quadrant (Q2) events with the largest contribution to the mean Reynolds shear stress were identified. The change in the inclination angle of Q2 events varies with wall units in $y^+<50$ and with the channel half height in y/h > 0.5. Conditionally averaged flow fields for the Q2 event show that the flow structures associated with Reynolds shear stress are a quasi-streamwise vortex in the buffer layer and a hairpin-shaped vortex in the outer layer. Three-dimensional visualization of the distribution of high Reynolds shear stress reveals that the organization of hairpin vortices in the outer layer having a size of 1.5~3 h is associated with large-scale motions with high Reynolds shear stress in the outer layer.