• Journal of Digital Convergence
• /
• v.18 no.11
• /
• pp.259-266
• /
• 2020

Many developments have been steadily carried out by researchers with applying knowledge-based expert system or machine learning algorithms to the financial field. In particular, it is now common to perform knowledge based system trading in using stock prices. Recently, deep learning technologies have been applied to real fields of stock trading marketplace as GPU performance and large scaled data have been supported enough. Especially, LSTM has been tried to apply to stock price prediction because of its compatibility for time series data. In this paper, we implement stock price prediction using LSTM. In modeling of LSTM, we propose a fitness combination of model parameters and activation functions for best performance. Specifically, we propose suitable selection methods of initializers of weights and bias, regularizers to avoid over-fitting, activation functions and optimization methods. We also compare model performances according to the different selections of the above important modeling considering factors on the real-world stock price data of global major companies. Finally, our experimental work brings a fitness method of applying LSTM model to stock price prediction.

• Microbiology and Biotechnology Letters
• /
• v.32 no.2
• /
• pp.135-141
• /
• 2004

Human granulocyte colony-stimulating factor (hG-CSF) is a hematopoiesis agent that principally affects the differentiation of neutrophils in the bone marrow. At present, recombinant hG-CSF is used successfully in the treatment of chemotheraphy-induced neutropenia and its indication has been expanded to bone marrow transplantation and aplastic anemia. In this study, we have constructed rhG-CSF secretion plasmid pYRC1 in which OmpA signal sequence/hG-CSF gene was expressed under the control of the T7 promoter. rhG-CSF produced in E. coli BL21 (pYRC1) grown at $37{\circ}C$ was found in aggregates. However, 15% of the periplasmic protein was soluble rhG-CSF when the E. coli BL21 (pYRC1) was cultured at $25^{\circ}C$ for 7 h in the modified MBL medium containing 10 g/$\ell$ glucose with 10 $\mu$M IPTG induction. The production of soluble rhG-CSF in E. coli BL21 (pYRC1) using fed batch culture was also studied. In the fed batch culture system, the final yield of rhG-CSF produced from E. coli BL21 (pYRC1) was increased from 4.4 mg/$\ell$to 24 mg/$\ell$by controlling the specific growth rate from $0.43 h^{-1}$ to $0.14 h^{-1}$, and optimizing the time of induction.

• Journal of Life Science
• /
• v.27 no.12
• /
• pp.1403-1409
• /
• 2017

In this study, the xylitol dehydrogenase (XYL2) gene was expressed in Saccharomyces cerevisiae as a host cell for ease of use in the degradation of lignocellulosic biomass (xylose). To select suitable expression systems for the S.XYL2 gene from S. cerevisiae and the P.XYL2 gene from Pichia stipitis, $pGMF{\alpha}-S.XYL2$, $pGMF{\alpha}-P.XYL2$, $pAMF{\alpha}-S.XYL2$ and $pAMF{\alpha}-P.XYL2$ plasmids with the GAL10 promoter and ADH1 promoter, respectively, were constructed. The mating factor ${\alpha}$ ($MF{\alpha}$) signal sequence was also connected to each promoter to allow secretion. Each plasmid was transformed into S. cerevisiae $SEY2102{\Delta}trp1$ strain and the xylitol dehydrogenase activity was investigated. The GAL10 promoter proved more suitable than the ADH1 promoter for expression of the XYL2 gene, and the xylitol dehydrogenase activity from P. stipitis was twice that from S. cerevisiae. The xylitol dehydrogenase showed $NAD^+$-dependent activity and about 77% of the recombinant xylitol dehydrogenase was secreted into the periplasmic space of the $SEY2102{\Delta}trp1/pGMF{\alpha}-P.XYL2$ strain. The xylitol dehydrogenase activity was increased by up to 41% when a glucose/xylose mixture was supplied as a carbon source, rather than glucose alone. The expression system and culture conditions optimized in this study resulted in large amounts of xylitol dehydrogenase using S. cerevisiae as the host strain, indicating the potential of this expression system for use in bioethanol production and industrial applications.

• Journal of the Korean Vacuum Society
• /
• v.7 no.4
• /
• pp.403-409
• /
• 1998

In this study, the effects of multi-dipole type of magnets on the characteristics of the inductively coupled plasmas and $SiO_2$ etch properties were investigated. As the magnets, 4 pairs of permanent magnets were used and, to etch $SiO_2, C_2F_6, CHF_3, C_4F_8, H_2$, and their combinations were used. The characteristics of the magnetized inductively coupled plasmas were investigated using a Langmuir probe and an optical emission spectrometer, and $SiO_2$ etch rates and the etch selectivity over photoresist were measured using a stylus profilometer. The use of multi-dipole magnets increased the uniformity of the ion density over the substrate location even though no significant increase of ion density was observed with the magnets. The use of the magnets also increased the electron temperature and radical densities while reducing the plasma potential. When $SiO_2$ was etched using the fluorocarbon gases, the significant increase of $SiO_2$ etch rates and also the increase of etch uniformity over the substrate were obtained using the magnets. In case of gas combinations with hydrogen, $C_4F_8/H_2$ showed the highest etch rates and etch selectivities over photoresist among the gas combinations with hydrogen used in the experiment. By optimizing process parameters at 1000 Watts of inductive power with the magnets, the highest $SiO_2$ etch rate of 8000 $\AA$/min could be obtained for 50% $C_4F_8/50% H_2$.

• The Journal of the Acoustical Society of Korea
• /
• v.42 no.6
• /
• pp.536-543
• /
• 2023

Speaker diarization, which labels for "who spoken when?" in speech with multiple speakers, has been studied on a deep neural network-based end-to-end method for labeling on speech overlap and optimization of speaker diarization models. Most deep neural network-based end-to-end speaker diarization systems perform multi-label classification problem that predicts the labels of all speakers spoken in each frame of speech. However, the performance of the multi-label-based model varies greatly depending on what the threshold is set to. In this paper, it is studied a speaker diarization system using single-label classification so that speaker diarization can be performed without thresholds. The proposed model estimate labels from the output of the model by converting speaker labels into a single label. To consider speaker label permutations in the training, the proposed model is used a combination of Permutation Invariant Training (PIT) loss and cross-entropy loss. In addition, how to add the residual connection structures to model is studied for effective learning of speaker diarization models with deep structures. The experiment used the Librispech database to generate and use simulated noise data for two speakers. When compared with the proposed method and baseline model using the Diarization Error Rate (DER) performance the proposed method can be labeling without threshold, and it has improved performance by about 20.7 %.

• Journal of Korea Water Resources Association
• /
• v.39 no.8 s.169
• /
• pp.677-690
• /
• 2006

The LH-OAT (Latin Hypercube One factor At a Time) method for sensitivity analysis and SCE-UA (Shuffled Complex Evolution at University of Arizona) optimization method were applied for the automatic calibration of SWAT model in Bocheong-cheon watershed. The LH-OAT method which combines the advantages of global and local sensitivity analysis effectively identified the sensitivity ranking for the parameters of SWAT model over feasible parameter space. Use of this information allows us to select the calibrated parameters for the automatic calibration process. The performance of the automatic calibration of SWAT model using SCE-UA method depends on the length of calibration period, the number of calibrated parameters, and the selection of statistical error criteria. The performance of SWAT model in terms of RMSE (Root Mean Square Error), NSEF (Nash-Sutcliffe Model Efficiency), RMAE (Relative Mean Absolute Error), and NMSE (Normalized Mean Square Error) becomes better as the calibration period and the number of parameters defined in the automatic calibration process increase. However, NAE (Normalized Average Error) and SDR (Standard Deviation Ratio) were not improved although the calibration period and the number of calibrated parameters are increased. The result suggests that there are complex interactions among the calibration data, the calibrated parameters, and the model error criteria and a need for further study to understand these complex interactions at various representative watersheds.

• Journal of Korean Tunnelling and Underground Space Association
• /
• v.15 no.3
• /
• pp.357-373
• /
• 2013

In this paper, a back analysis program for analyzing the behavior of tunnel-ground system and evaluating the material properties and tunnel design parameters was developed. This program was designed to be able to implement the back analysis of underground structure by combination of using FLAC and optimized algorithm as direct method. In particular, Rosenbrock method which is able to do direct search without obtaining differential coefficient was adopted for the back analysis algorithm among optimization methods. This back analysis program was applied to the site to evaluate the design parameters. The back analysis was carried out using field measurement results from 5 sites. In the course of back analysis, nonlinear regression analysis was carried out to identify the optimum function of the measured ground displacement. Exponential function and fractional function were used for the regression analysis and total displacement calculated by optimum function was used as the back analysis input data. As a result, displacement recalculated through the back analysis using measured displacement of the structure showed 4.5% of error factor comparing to the measured data. Hence, the program developed in this study proved to be effectively applicable to tunnel analysis.

• Journal of Korean Society of Environmental Engineers
• /
• v.30 no.8
• /
• pp.839-846
• /
• 2008

The flocculation characteristics of polyacrylamide base flocculants were estimated to reduce the moisture content of the dehydrated cakes. The dewaterability for sewage sludge was found to have a marked effect depending on the flocculant type, agitating speed and time, kind of dissolution water, etc. The optimal agitating speed and time were 700 rpm and 3 sec, respectively, in this experimental condition. and the dewaterability was proportion to the agitating speed upto 700rpm. When recycle water as the dissolution water was used, the solution viscosity of all kind of flocculants was decreased. However, the change of its viscosity are not proportioned to the dewaterabilities for each flocculant. Flocculation system of combinations of the first and sencond flocculation using single flocculant was investigated. Effects of the ratio of first and second dosage for dual flocculation on the dewaterability were also investigated. The optimum conditions of dual flocculation system are 75% and 50% as first dosages for low and high viscous flocculant for total dosage of common flocculation, respectively. Based on the results, an overall mechanism of dual flocculation system is proposed and it is envisaged that optimization of flocculation processes in this way can result in considerable savings in cost.

• Journal of the Korean Ceramic Society
• /
• v.39 no.6
• /
• pp.582-588
• /
• 2002

IrO$_2$-RuO$_2$ films were deposited on plasma sprayed TiO$_2$ buffer layer above Ti metal by sol-gel and dip-coating method. Organic vehicle (ethyl cellulose and $\alpha$-terpineol) and glass frit were added to improve adherence of the coatings. Taguchi method and L$_{18}$ (2$^1$$\times$3$^{7}$ ) orthogonal arrays were evalvated in terms of current density to determine the optimal combination of levels of factors that best satisfy the bigger is better quality characteristic. The observed conditions were as fellows: ethyl cellulose (100 cp), drying temperature and time (17$0^{\circ}C$,20 min), heat treatment temperature and time (75$0^{\circ}C$,10 min), the weight ratio of IrO$_2$-RuO/powders to glass frit (99:5), final heat treatment time (120 min) and flow rate of air (5 sccm), respectively. ANOVA analysis suggested that the influence of the factors within $\alpha$= 0.1 was significant with a 90% confidence level.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.13 no.2
• /
• pp.200-208
• /
• 2003

This paper proposes a separation and feature extraction of image signals using the independent component analysis(ICA) based on neural networks of efficient learning rule. The proposed learning rule is a hybrid fixed-point(FP) algorithm based on secant method and momentum. Secant method is applied to improve the performance by simplifying the 1st-order derivative computation for optimizing the objective function, which is to minimize the mutual informations of the independent components. The momentum is applied for high-speed convergence by restraining the oscillation in the process of converging to the optimal solution. The proposed algorithm has been applied to the composite images generated by random mixing matrix from the 10 images of $512\times512$-pixel. The simulation results show that the proposed algorithm has better performances of the separation speed and rate than those using the FP algorithm based on Newton and secant method. The proposed algorithm has been also applied to extract the features using a 3 set of 10,000 image patches from the 10 fingerprints of $256\times256$-pixel and the front and the rear paper money of $480\times225$-pixel, respectively, The simulation results show that the proposed algorithm has also better extraction speed than those using the another methods. Especially, the 160 basis vectors(features) of $16\times16$-pixel show the local features which have the characteristics of spatial frequency and oriented edges in the images.