• Journal of Intelligence and Information Systems
• /
• v.23 no.2
• /
• pp.1-17
• /
• 2017

The deep learning framework is software designed to help develop deep learning models. Some of its important functions include "automatic differentiation" and "utilization of GPU". The list of popular deep learning framework includes Caffe (BVLC) and Theano (University of Montreal). And recently, Microsoft's deep learning framework, Microsoft Cognitive Toolkit, was released as open-source license, following Google's Tensorflow a year earlier. The early deep learning frameworks have been developed mainly for research at universities. Beginning with the inception of Tensorflow, however, it seems that companies such as Microsoft and Facebook have started to join the competition of framework development. Given the trend, Google and other companies are expected to continue investing in the deep learning framework to bring forward the initiative in the artificial intelligence business. From this point of view, we think it is a good time to compare some of deep learning frameworks. So we compare three deep learning frameworks which can be used as a Python library. Those are Google's Tensorflow, Microsoft's CNTK, and Theano which is sort of a predecessor of the preceding two. The most common and important function of deep learning frameworks is the ability to perform automatic differentiation. Basically all the mathematical expressions of deep learning models can be represented as computational graphs, which consist of nodes and edges. Partial derivatives on each edge of a computational graph can then be obtained. With the partial derivatives, we can let software compute differentiation of any node with respect to any variable by utilizing chain rule of Calculus. First of all, the convenience of coding is in the order of CNTK, Tensorflow, and Theano. The criterion is simply based on the lengths of the codes and the learning curve and the ease of coding are not the main concern. According to the criteria, Theano was the most difficult to implement with, and CNTK and Tensorflow were somewhat easier. With Tensorflow, we need to define weight variables and biases explicitly. The reason that CNTK and Tensorflow are easier to implement with is that those frameworks provide us with more abstraction than Theano. We, however, need to mention that low-level coding is not always bad. It gives us flexibility of coding. With the low-level coding such as in Theano, we can implement and test any new deep learning models or any new search methods that we can think of. The assessment of the execution speed of each framework is that there is not meaningful difference. According to the experiment, execution speeds of Theano and Tensorflow are very similar, although the experiment was limited to a CNN model. In the case of CNTK, the experimental environment was not maintained as the same. The code written in CNTK has to be run in PC environment without GPU where codes execute as much as 50 times slower than with GPU. But we concluded that the difference of execution speed was within the range of variation caused by the different hardware setup. In this study, we compared three types of deep learning framework: Theano, Tensorflow, and CNTK. According to Wikipedia, there are 12 available deep learning frameworks. And 15 different attributes differentiate each framework. Some of the important attributes would include interface language (Python, C ++, Java, etc.) and the availability of libraries on various deep learning models such as CNN, RNN, DBN, and etc. And if a user implements a large scale deep learning model, it will also be important to support multiple GPU or multiple servers. Also, if you are learning the deep learning model, it would also be important if there are enough examples and references.

• Journal of the Korean Institute of Traditional Landscape Architecture
• /
• v.33 no.4
• /
• pp.38-51
• /
• 2015

In the maps of the period, there were three large ponds called Dongji(東池), Seoji(西池) and Namji(南池) in Hanyang, the capital of Joseon Dynasty. They were different than the ponds found in the palace, civic buildings, and private dwellings. Dongji, Seoji and Namji were ponds relating to Fortress wall of Seoul, and all had lotuses cultivated in them. The purpose of this paper is to clarify the locational and spatial characteristics of these ponds and to detail the construction and reconstruction process and management conditions through maps, drawings, illustrations, historical records and literary works from the urban environmental perspective. The results are as follows. First, Seoji and Namji were intended for Bibo(裨補) which redeemed the geographical weaknesses of Hanyang, securement of bright court water(明堂水), supplement for fire energy(火氣), fire preventive water and waterscape facilities, while Dongji was emphasized on protecting water mouth(水口) besides Bibo and securement of bright court water. Second, Seoji was connected to mountain streams and Dongji and Namji were to ditches. The ponds connected to ditches had been difficult to fill and maintain. Third, Seoji and Namji were in urban areas, whereas Dongji was in farmlands, and these locational differences had an influence on the use of ponds. Fourth, the shapes of ponds, in contrast to the ponds in palace and civic buildings, which were perfectly square, were either freeform or square with rounded edges. Fifth, lotus ponds could be maintained by continuous management polices, earth filling and reconstructing process were repeated during the Joseon Dynasty. The lotus ponds of Fortress Wall of Seoul which had managed over 500 years, were built in, in accordance with the tenets of Bibo pungsu geomancy; however as time passed, they were maintained not only as public open spaces, but also a cultural attraction for residents and visitors.

• The korean journal of orthodontics
• /
• v.28 no.1 s.66
• /
• pp.85-97
• /
• 1998

The purpose of this study was to observe the effects of sodium fluoride on the bony repair and regeneration processes after the rapid palatal expansion in the growing dogs. Eighteen dogs were divided into experimental and control groups. They were in the late mixed dentition. The rapid Palatal expansion was undertaken in all the animals($180^{\circ}$ turn/day) for ten days. The animals were sacrificed on 0, 15 and 45 days after the finish of expansion. One mg NaF/kg of body weight/day were given orally to the experimental group. Blood samples were drawn before and after expansion and the se겨m calcium, phosphate and alkaline phosphatase level were measured. The undecalcified bone section of midpalatal suture area was made, and observed under the light microscopy The results were as follows ; 1. The day after expansion, the infiltration of inflammatory cells were prominent and the new bone formation started at the edges of the two palatal plates bodering the midpalatal suture in both groups. Especially, the newly formed osteoid were very extensive and the osteoblasts lining the osteoid were very active in the experimental group. 2. At fifteen days after expansion, the active osteoblasts lining the osteold at the surface of trabecular bony spicules and active new bone formation were observed in the both groups. However, the cellular activity and new bone formation were more prominent In the experimental group. 3. At forty five days after expansion, the continuous osteoid and new bone formation and active osteoblasts were observed in the experimental group. But these phenomena were not observed in the control group. In the control group, the numerous osteoclasts were adjacent midpalatal suture and the bony remodeling process was begun. The serum alkaline phosphatase level was maintained highly in the experimental group, but decreased in the control. According to the above results, the author reached the conclusion that sodium fluoride has the stimulation effects on the osteoid production of the osteoblasts during the healing process after the rapid Palatal expansion more continuously.

• Journal of the Korean Geographical Society
• /
• v.30 no.3
• /
• pp.296-310
• /
• 1995

Classification of spatial units into meaningful sets is an important procedure in spatial analysis. It is crucial in characterizing and identifying spatial structures. But traditional classification methods such as cluster analysis require an exact database and impose a clear-cut boundary between classes. Scrutiny of realistic classification problems, however, reveals that available infermation may be vague and that the boundary may be ambiguous. The weakness of conventional methods is that they fail to capture the fuzzy data and the transition between classes. Fuzzy subsets theory is useful for solving these problems. This paper aims to come to the understanding of theoretical foundations of fuzzy spatial analysis, and to find the characteristics of fuzzy classification methods. It attempts to do so through the literature review and the case study of urban classification of the Cities and Eups of Kyung-Nam Province. The main findings are summarized as follows: 1. Following Dubois and Prade, fuzzy information has an imprecise and/or uncertain evaluation. In geography, fuzzy informations about spatial organization, geographical space perception and human behavior are frequent. But the researcher limits his work to numerical data processing and he does not consider spatial fringe. Fuzzy spatial analysis makes it possible to include the interface of groups in classification. 2. Fuzzy numerical taxonomic method is settled by Deloche, Tranquis, Ponsard and Leung. Depending on the data and the method employed, groups derived may be mutually exclusive or they may overlap to a certain degree. Classification pattern can be derived for each degree of similarity/distance $\alpha$. By takina the values of $\alpha$ in ascending or descending order, the hierarchical classification is obtained. 3. Kyung-Nam Cities and Eups were classified by fuzzy discrete classification, fuzzy conjoint classification and cluster analysis according to the ratio of number of persons employed in industries. As a result, they were divided into several groups which had homogeneous characteristies. Fuzzy discrete classification and cluste-analysis give clear-cut boundary, but fuzzy conjoint classification delimit the edges and cores of urban classification. 4. The results of different methods are varied. But each method contributes to the revealing the transparence of spatial structure. Through the result of three kinds of classification, Chung-mu city which has special characteristics and the group of Industrial cities composed by Changwon, Ulsan, Masan, Chinhai, Kimhai, Yangsan, Ungsang, Changsungpo and Shinhyun are evident in common. Even though the appraisal of the fuzzy classification methods, this framework appears to be more realistic and flexible in preserving information pertinent to urban classification.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.14 no.2
• /
• pp.157-167
• /
• 2003

In this paper, to reduce the resonant length of patch, microstrip patch antenna of linear polarization which is suppressed at two radiation edges is designed and fabricated at the frequency of 1.575 GHz. The result is like that the resonant length of patch is 45 mm and the length reduction effect is 43.8 % when it is compared with that(80 mm) of plane type. The gain is 4.4 dBd and -3 dB beamwidths are 112$^{\circ}$ and 66$^{\circ}$ in the E-plane and H-plane, respectively. Also, to reduce the size of patch, microstrip patch antennas those are suppressed at four radiating comers are designed and fabricated at the same frequency in the linear and circular polarization, respectively. For linear polarization, at the 1.2 of width/length(W/L) ratio, the patch area is 53 mm $\times$ 63.6 mm and the size reduction effect is 56.1 % when compared with that(80 mm $\times$ 96 mm) of plane type. The gain is 4.3 dBd and the -3 dB beamwidths are 120$^{\circ}$ and 78$^{\circ}$ in the E-plane and H-plane, respectively. For circular polarization, the patch size(54.2 mm $\times$ 61.5 mm) is reduced by 47.2 % than that(76 mm $\times$ 83 mm) of plane type. -3 dB beamwidth of horizontal polarization in the z-x plane and vortical polarization in the y-z plane are 108$^{\circ}$ and 93$^{\circ}$, respectively and this means the increasement in both planes by 52$^{\circ}$ and 27$^{\circ}$ than those of plane type. The maximum gain is 2.5 dBd in the horizontal polarization in the z-x plane. Axial ratio is 1.5 dB at 1.575 GHz and the 2 dB axial ratio bandwidth(ARBW) is 20 MHz(1.3 %).

• Economic and Environmental Geology
• /
• v.38 no.2 s.171
• /
• pp.187-196
• /
• 2005

Structural sites which cations can occupy in garnet structure are centers of the tetrahedron, octahedron, and distorted cube sharing edges with the tetrahedron and octahedron. Among them, the size of cation occuping at tetrahedral site (the center of tetrahedron) is closely related with the size of a unit cell of garnet. Accordingly, garnet containing iron with relative large ionic radii in tetrahedral site can be considered as a promising matrix for the immobilization of the elements with large ionic radii, such as actinides in radioactive wastes. We synthesized several garnets with the batch composition of $Ca_{1.5}GdCe_{0.5}ZrFeFe_3O_{12}$, and studied their properties and phase relations under various conditions. Mixed samples were fabricated in a pellet form under a pressure of $200{\~}400{\cal}kg/{\cal}cm^2$ and were sintered in the temperature range of $1100\~1400^{\circ}C$ in air and under oxygen atmospheres. Phase identification and chemical analysis of synthesized samples were conducted by XRD and SEM/EDS. In results, garnet was obtained as the main phase at $1300^{\circ}C$, an optimum condition in this system, even though some minor phases like perovskite and unknown phase were included. The compositions of garnet and perovskite synthesized from the batch composition of $Ca_{1.5}GdCe_{0.5}ZrFeFe_3O_{12}$ were ranged $[Ca_{l.2-1.8}Gd_{0.9-1.4}Ce_{0.3-0.5}]^{VIII}[Zr_{0.8-1.3}Fe_{0.7-1.2}]^{VI}[Fe_{2.9-3.1}]^{IV}O_{12}$ and $Ca_{0.1-0.5}Gd_{0.0-0.8}Ce_{0.1-0.5}\;Zr_{0.0-0.2}Fe_{0.9-1.1}O_3$, respectively. Ca content was exceeded and Ce content was depleted in the 8-coordinated site, comparing to the initial batch composition. This phenomena was closely related to the content of Zr and Fe in the 6-coordinated site.

• Journal of Fashion Business
• /
• v.3 no.1
• /
• pp.65-80
• /
• 1999

In this study, the researcher studied the historical background and the traditional culture about dress and ornament of Yunnan Province of China. The Results of the study are as follows. 1. In the Past, Derung's dress was very simple due to the influence of various factors, such as geography and history. Men wore shorts and covered diagonally a piece of cloth from left shoulders to right armpits and tied up the two ends on chests. Women covered crisscross two pieces of cloth from both shoulders to knee. 2. Achang people's dress and adornment has its own unique characteristic. Generally, men wear Jackets with buttons down the front and black trousers. Unmarried men like to wear white turbans, while most of married men usually wear dark blue ones. Women usually wear tight-sleeve blouses with buttons down the front and skirts. Unmarried women wear the hair in braids coil them on the top of their heads. They wear short blouses and trousers. Married women wear their hair Into buns and like to entwine black or blue cloth into high trubans. They wear short blouses and knee-length straight skirts. Achang knife enjoys high reputation and has a long history and an exquisite workmanship. All the men like to wear it. 3. The dress and adornment of the Lahu nationality has both the characteristic of farming culture and the style of nomadic culture of early times. Men usually wear short shirts with round necks and buttons down the front, loose-legged trousers, turbans or dark blue cloth caps Women's dress and adornment can be categorized into two styles. One is black cloth gown with buttons diagonally on the right front and waist-length slits on both sides. The edges of fronts and cuffs are edged with Silver ornaments and lace. They also wear trousers. The other is short blouse with round neck and short opening on th right front, straight skirt and colourful leggings with embroidered patterns. 4. The Hani people, men and women, old and young, like black colour and are fond of wearing black clothes. Men usually wear shirts with buttons down the front and trousers, entwining their heads with black or white cloth. The elderly people wear calottes. Women wear cloth blouses, skirts and trousers or shorts. Slight differences exist in the clothing and adornments according to region, branch and age 5. Blang people's dyeing technique with an exquisite method has a long history. Men wear dark blue long sleeve shirts with round necks and buttons down the front or arranged diagonally on the front and loose-legged trousers. Elderly men wear big turbans wdress and adornment varies greatly in different regions. 6. The Lisu people culture of dress and adornment has some unique characteristics. The styles and colours of their dress and adornment differ slightly from place to place. In the Nujiang area, Women wear black velvet Jackets over blouses with buttons arranged diagonally on the right front and long pleated ramie skirts. Men usually wear wraparound ramie gowns, with center vent, made of fabrics alternated with white and black cross stripes. They also wear cloth waistbands and trohile youngsters keep their hair short. Women's users. In the Lushui area, the dress and adornment is similar to that in the Nujing region, but women wear aprons and trousers instead of skirts. 7. The Nu people dress and adornment is simple but elegant Women are proficient in ramie-weaving. Men usually wear gowns With overlapping necks, knee-length trousers and leggings. They like to wear their hair long and entwine dark blue or white turbans. Women wear black and red vests over blouses with buttons arranged diagonally on the right front and ankle-length skirts. They also wear their hair long, make it into braids, and entwine dark blue or colourful cloth turbans. 8. Pumi men usually wear ramie shirts With buttons arranged diagonally on the right front, loose trousers and white sheepskin vests. Some also wear overcoats made of "pulu". Women's dress and adornment varies in different areas. In the Lanping and Weixi regions, women wear white short blouses with buttons arranged diagonally on the front and dark brown embroidered vests. They also wear trousers and blue or black cloth turbans. In the Ninglang and Yongsheng regions, women wear hemmed blouses With buttons arranged diagonally on the right front and drape sheepskin capes. They also wear white pleated skirts and use broad colourful cloth as their waistbands.

• Journal of Korean Neurosurgical Society
• /
• v.30 no.3
• /
• pp.263-271
• /
• 2001

Objective : Glioblastomas, the most common type of primary brain tumors, are highly invasive and cause massive tissue destruction at both the tumor invading edges and in areas that are not in direct contact with glioma cells. As a result, patients with high-grade gliomas are faced with a poor prognosis. Such grim statistics emphasize the need to better understand the mechanisms that underlie glioma invasion, as these may lead to the identification of novel targets in the therapy of high grade gliomas. Protein kinase C(PKC) is a family of serine/threonine kinases and an important signal transduction enzyme that conveys signals generated by ligand-receptor interaction at the cell surface to the nucleus. PKC appears to be critical in regulating many aspects of glioma biology. The purpose of this study was to assess accurately the role of PKC in the invasion regulation of human gliomas based on hypothesis that protein kinase C(PKC) is functional in the process of glial tumor cell invasion. Method : To test this hypothesis, U-87 malignant glioma cell line intracellular PKC levels were up and down regulated and their invasiveness was tested. Intracellular PKC level was characterized using PKC activity assays. Invasion assays including barrier migration and spheroid confrontation were used to study the relationship between PKC concentration and invasiveness. Result : The cell line which were treated by PKC inhibitor tamoxifen and hypericin exhibited decreased PKC activity and decreased invasive abilities dose dependently both in matrigel invasion assay and tumor spheroid fetal rat brain aggregates(FRBA) confrontation assay. However, the cell line that was treated by PKC activator 12-O-tetradecanylphorbol-13acetate(TPA) did not exhibit increases in either PKC activity or invasive ability. Conclusion : These studies suggest that PKC may be a useful molecular target for the chemotherapy of glioblastoma and other malignancies and that a therapeutic approach based on the ability of PKC inhibitors may be helpful in preventing invasion.

• Journal of Korean Society of Environmental Engineers
• /
• v.33 no.12
• /
• pp.907-912
• /
• 2011

Development of visible light responsive photocatalysts is a promising research area to facilitate utilization of solar energy for hydrogen production via photocatalytic water splitting. In this study two groups of samples, nitrogen (N)-doped niobium pentoxide ($Nb_2O_5$) and titanium dioxide ($TiO_2$) ($Nb_2O_5-N$, $HNb_3O_8-N$, $TiO_2-N$) and N-undoped ones ($Nb_2O_5$ and $TiO_2$) were tested. In order to utilize visible light, nitrogen atoms were doped in selected photocatalysts by using urea. A shift of the absorption edges of the Ndoped samples in the visible light region was observed. Under visible light irradiation, N-doped samples were more prominent photocatalytic activities than the N-undoped samples. Specifically, 99.7% of rhodamine B (RhB) was degraded after 60 minutes of visible light irradiation with $TiO_2-N$. Since $TiO_2-N$ shows the highest activity of RhB degradation, it was supposed to generate the highest current response. However, $HNb_3O_8-N$ showed the highest current response ($63.7mA/cm^2$) than $TiO_2-N$. More interestingly, when we compare the hydrogen production, $Nb_2O_5-N$ produced $19.4{\mu}mol/h$ of hydrogen.

• The korean journal of orthodontics
• /
• v.30 no.3 s.80
• /
• pp.317-333
• /
• 2000

The purpose of this study was to determine the proliferative activity of the osteoblasts and fibroblasts in the midpalatal area and to investigate the adjacent periodontal tissues of individual tooth following rapid expansion of the palate. Ten young adult dogs, aged approximately ten months, were used in the experiment. The experimental design was consisted of 1 week expansion group(Group E1, 3 dogs), 2 week expansion group(Group E2, 3 dogs), 2 week expansion and 2 week retention group(Group E3, 3 dogs), and control group(Group C, 1 dog). For each group, expansion screw was activated one time per day(1/4 turn;$90^{\circ}$) following Hyrax-screw application. The experimental animals in each group were sacrificed at 1, 2 and 4 weeks following palatal expansion. Maxillary tissue blocks were obtained and prepared ior the histomorphologic and immunohistochemical studies. Light mcroscope, polarizing microscope, and soft X-ray apparatus were used in this study, and following results were obtained. 1. In polarizing microscopic study, the expansion groups(E1 & E2) showed blue color representing bone resorption and new bone formation in midpalatal suture area. E3 groups skewed less blue color compared to the E1 and E2 group. But yellow color increased by calcification in the E3 groups. 2. Immunohistochemical study revealed that positive responses of the osteoblasts to PCNA and undifferentiated fibroblasts to EGF in E1 group were somewhat increased. Positive response to PCNA and EGF were increased in fibroblasts and the osteoblasts forming new bone in E2 group. In E3 group, the positive response cell concentrated the periphery of edge of palatal process in both PCNA and EGF. 3. Throughout the expansion period(E1 & E2), light microscopic study showed the edges of the extensive resorption and new palatal processes, indicating bone remodeling within the suture. E3 group exhibited less remodeling of midpalatal suture area. E2 group and E3 group showed cementum formation and resorption at the apex of 3rd premolar and 1st molar E3 group exhibited extensive hyalinized zone on the cervical portion of buccal side of 1st molar. 4. Soft X-ray analysis of E1 group showed hypomineralized defect and microfractures in various parts of the suture areas when compared with control animals. There was no significant difference in the degree of mineralization in the midpalatal suture region between the C and E3 groups. Tooth axis showed tipping of 3rd premolar and 1st molar in the E2 group and E3 group. Based upon these experimental results, it is concluded that the undifferentiated mesenchymal cells always presented in midpalatal suture area following RPE. Differentiated osteoblasts and fibroblasts possess proliferating cellular activity until the 2 week retention period. The posterior teeth are tend to tip buccally as RPE force applied. Retention group exhibited irreversible response with severe hyalinized zone on the buccal surface of the first molar.