• Journal of Intelligence and Information Systems
• /
• v.26 no.4
• /
• pp.27-65
• /
• 2020

Many companies on information and communication technology make public their own developed AI technology, for example, Google's TensorFlow, Facebook's PyTorch, Microsoft's CNTK. By releasing deep learning open source software to the public, the relationship with the developer community and the artificial intelligence (AI) ecosystem can be strengthened, and users can perform experiment, implementation and improvement of it. Accordingly, the field of machine learning is growing rapidly, and developers are using and reproducing various learning algorithms in each field. Although various analysis of open source software has been made, there is a lack of studies to help develop or use deep learning open source software in the industry. This study thus attempts to derive a strategy for adopting the framework through case studies of a deep learning open source framework. Based on the technology-organization-environment (TOE) framework and literature review related to the adoption of open source software, we employed the case study framework that includes technological factors as perceived relative advantage, perceived compatibility, perceived complexity, and perceived trialability, organizational factors as management support and knowledge & expertise, and environmental factors as availability of technology skills and services, and platform long term viability. We conducted a case study analysis of three companies' adoption cases (two cases of success and one case of failure) and revealed that seven out of eight TOE factors and several factors regarding company, team and resource are significant for the adoption of deep learning open source framework. By organizing the case study analysis results, we provided five important success factors for adopting deep learning framework: the knowledge and expertise of developers in the team, hardware (GPU) environment, data enterprise cooperation system, deep learning framework platform, deep learning framework work tool service. In order for an organization to successfully adopt a deep learning open source framework, at the stage of using the framework, first, the hardware (GPU) environment for AI R&D group must support the knowledge and expertise of the developers in the team. Second, it is necessary to support the use of deep learning frameworks by research developers through collecting and managing data inside and outside the company with a data enterprise cooperation system. Third, deep learning research expertise must be supplemented through cooperation with researchers from academic institutions such as universities and research institutes. Satisfying three procedures in the stage of using the deep learning framework, companies will increase the number of deep learning research developers, the ability to use the deep learning framework, and the support of GPU resource. In the proliferation stage of the deep learning framework, fourth, a company makes the deep learning framework platform that improves the research efficiency and effectiveness of the developers, for example, the optimization of the hardware (GPU) environment automatically. Fifth, the deep learning framework tool service team complements the developers' expertise through sharing the information of the external deep learning open source framework community to the in-house community and activating developer retraining and seminars. To implement the identified five success factors, a step-by-step enterprise procedure for adoption of the deep learning framework was proposed: defining the project problem, confirming whether the deep learning methodology is the right method, confirming whether the deep learning framework is the right tool, using the deep learning framework by the enterprise, spreading the framework of the enterprise. The first three steps (i.e. defining the project problem, confirming whether the deep learning methodology is the right method, and confirming whether the deep learning framework is the right tool) are pre-considerations to adopt a deep learning open source framework. After the three pre-considerations steps are clear, next two steps (i.e. using the deep learning framework by the enterprise and spreading the framework of the enterprise) can be processed. In the fourth step, the knowledge and expertise of developers in the team are important in addition to hardware (GPU) environment and data enterprise cooperation system. In final step, five important factors are realized for a successful adoption of the deep learning open source framework. This study provides strategic implications for companies adopting or using deep learning framework according to the needs of each industry and business.

• Journal of Korean Society of Steel Construction
• /
• v.17 no.5 s.78
• /
• pp.561-568
• /
• 2005

The object of this paper is to determine appropriateness for use of high-strength tensile bar as a tension web member. The gap K-joint of tensile bar types were compared with gap K-joint of square hollow section (SHS) types. For the same width-to-thickness ratio ($2{\gamma}=33.3$ ), tests were performed on four specimens of the SHS type and eight specimens of the tensile bar type. The comparison of capacity with the experimental results showed a capacity of the SHS type joint to be higher than that of the tensile bartype joint for the same brace-to-chord width ratio. Moreover, the capacity of the SHS type joints increased proportionally to the width ratio ${\beta}$), while tensile bar type joints increased as the tension width ratio (${\beta}2$). In failure mode, SHS-type specimens showed local buckling of the compression brace and plastic failure was observed between the tension brace and chord face, and with the tensile bar type specimens there appeared punching shear failure of the chord face at the toe of the connection plate. It is, therefore, concluded that width-to-thickness ratio should be lower than that of the hollow-section type and the relation between tension and compression width ratio should be considered.

• Journal of the Earthquake Engineering Society of Korea
• /
• v.2 no.4
• /
• pp.23-30
• /
• 1998

During an earthquake, there are three components of excitation : horizontal excitation of the ground, vertical excitation of the pile due to superstructure feedback produced by vertical excitation of the ground, and the seawater excitation by the vertical ground shaking, that is, "the seaquake." These excitations could have effects on the pore pressure buildup mode induced in the near-field of open-ended pile and the soil plugs in open-ended pipe piles installed at offshore sites. While the ground and pile excitation could be modeled by exciting the soil and pile with simulated motions, seaquake excitation induced by the vertical ground shaking can be modeled by pulsing the water pressure at the seabed. The objectives of this study were to observe buildup trend for the porewater pressures developed in near-field of open-ended pipe pile installed in the calibration chamber during the simulated earthquake and seaquake and, also to confirm the cause for reduction of soil plugging according to pore pressure buildup. During the simulated horizontal seismic motion, there was no upward flow through soil plug because the similar magnitude of excess porewater pressure were occurred at the top and under the toe of soil plug. During the horizontal seismic motion, relatively higher hydraulic gradients caused upward flow in the soil plug and then the degradation of plugging resistance was about 20%. During seaquake, in the case of the open-ended pile installed in a deep sea with more than 220m of water depth, soil plug failed completely because of high upward hydraulic gradients through soil plug.soil plug.

• Korean Journal of Applied Biomechanics
• /
• v.26 no.2
• /
• pp.167-174
• /
• 2016

Objective: The purpose of this study was to determine the interrelationship between ranges of motion of the knee and ankle joints on the sagittal plane and the attenuation magnitude of impact shock at high frequency (9~20 Hz) in the support phase during downhill running. Method: Fifteen male heel-toe runners with no history of lower extremity injuries were recruited for this study (age, $25.07{\pm}5.35years$; height, $175.4{\pm}4.6cm$; mass, $75.8{\pm}.70kg$). Two uniaxial accelerometers were mounted to the tuberosity of tibia and sacrum, respectively, to measure acceleration signals. The participants were asked to run at their preferred running speed on a treadmill set at $0^{\circ}$, $7^{\circ}$, and $15^{\circ}$ downhill. Six optical cameras were placed around the treadmill to capture the coordinates of the joints of the lower extremities. The power spectrum densities of the two acceleration signals were analyzed and used in the transfer function describing the gain and attenuation of impact shock between the tibia and the sacrum. Angles of the knee and ankle joints on the sagittal plane and their angle ranges were calculated. The Pearson correlation coefficient was used to test the relationship between two variables, the magnitude of impact shock, and the range of joint angle under three downhill conditions. The alpha level was set at .05. Results: Close correlations were observed between the knee joint range of motion and the attenuation magnitude of impact shock regardless of running slopes (p<.05), and positive correlations were found between the ranges of motion of the knee and ankle joints and the attenuation magnitude of impact shock in $15^{\circ}$ downhill running (p<.05). Conclusion: In conclusion, increased knee flexion might be required to attenuate impact shock during downhill and level running through change in stride or cadence while maintaining stability, and strong and flexible ankle joints are also needed in steeper downhill running.

• Geophysics and Geophysical Exploration
• /
• v.16 no.3
• /
• pp.109-118
• /
• 2013

On and near the 23-m high earthen Cheongcheon dam in Boryeong City, Korea, short seismic refraction and surface-wave profiles were conducted using a 5-kg sledgehammer. From vertical and horizontal components of the seismic waves, near-surface P-wave velocities (${\nu}_p$) and S-wave velocities (${\nu}_s$) were derived by inverting first-arrival refraction times and dispersion curves of Rayleigh waves. Average ${\nu}_p$ and ${\nu}_s$ for the Jurassic sedimentary basement were determined to be 1650 and 950 m/s at a depth of 30 m directly beneath the dam and 1650 m/s and 940 m/s at a depth of 10 m at the toe of the dam, respectively. The dynamic Poisson's ratio for these strata were therefore in the range of 0.24 to 0.25, which is consistent with ratios for consolidated sedimentary strata. Near a 45-m borehole 152 m downstream from the dam crest, an SH tomogram indicates a refraction boundary with an average ${\nu}_s$ of 870 m/s at depths of 10 ~ 12 m. At this site, the overburden comprises the upper layer with relatively constant ${\nu}_p$ and ${\nu}_s$ around 500 and 200 m/s, respectively, and the lower layer in which both ${\nu}_p$ and ${\nu}_s$ increase with depth almost linearly. The dynamic Poisson's ratios for the overburden were in the range of 0.30 to 0.43.

• Journal of the Korean Geotechnical Society
• /
• v.27 no.10
• /
• pp.25-33
• /
• 2011

Suction buckets have been widely used for offshore structures such as anchors for floating facilities, and the foundations of offshore wind energy turbines. However, the design guidelines for suction buckets have not been clearly suggested. Therefore, this study performed the numerical analysis by using ABAQUS (2010) to evaluate bearing capacities and load-movement behaviors of the suction bucket in NC clay. For the numerical analysis, the depth ratio L/D (L=embedded length of skirt; D=diameter of a bucket) was varied from 0.25 to 1.0. The analysis results showed that the L/D ratio has a significant effect on the bearing capacity, and the vertical and horizontal capacities respectively increased by about 40% and 90%, when L/D ratio increased from 0.25 to 1.0. At the vertical loading, the bucket showed the similar failure mode with a deep foundation, so the shaft and toe resistances can be separately evaluated. At the horizontal loading, the bucket with L/D=O.25 showed the sliding failure mode and the bucket with $L/D{\geq}0.5$ showed the rotational failure mode.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.5 no.4
• /
• pp.1-8
• /
• 1985

Stability analysis of the embankment as to water level varation is the most important problem in the safety of the slope because the stress of embankment inside varies as to drawdown of seepage line. Especially when the water level is rapidly drawdown, because the flow direction of the free surface changes the toe of embankment, the factor of safety comes to small, therefore the embankment is dangered. For the purpose of studing these phenomena, the experimental models are built with sand in the laboratory. In the experimental consideration, the falling seepage line and the shape of failure are measured. This paper intends to study the failure slip surface, the relationship between the factor of safety and drawdown velocity, and hydraulic gradient. The results of the experimental study are summarized as follows; 1. Owing to the drawdown of free surface, sliding failure occurred in the upstream fill, the height of failure is 5~10, 9~15, and 13~21(cm) in each model. 2. In consideration of the distribution of pore water pressure Table-5 shows each factor of safety. In the relationship between the drawdown velocity and the factor of factor it's velocity should be limited to 0.21~0.28 (cm/sec), according to each models. In the relationship between the factor of safety and the hydraulic gradient within the upstream slope, it's gradient must be below 0.36~0.43.

• Korean Journal of Applied Biomechanics
• /
• v.18 no.1
• /
• pp.107-115
• /
• 2008

The purpose of this study was to investigate muscle activity and vertical ground reaction force(F) in children with Down syndrome(DS) during vertical jump. Six DS and one healthy child performed vertical jump. Four muscles(Biceps femoris, Rectus femoris, Tibialis anterior & Gastrocnemius) and F were analyzed. Gastrocnemius in DS showed lower muscle activity in a propulsive phase. Impulse during 0.3sec before toe-off in DS displayed lower value than that in the healthy child. The second peak of F in DS occurred later than that in the healthy child, so DS performed landing with their knee more flexed. The first and second peak of F and loading rate to the second peak of F in DS showed lower value than those in the healthy child. Therefore, DS might have lower ability to absorb the force while landing from a vertical jump.

• Korean Journal of Applied Biomechanics
• /
• v.18 no.2
• /
• pp.1-9
• /
• 2008

The purpose of this study was to investigate the effect of different obstacle heights on the plantar foot pressure during obstacle crossing. Sixteen healthy adults who had no musculoskeletal disorders were instructed to perform unobstructed level walking and to step over obstacles corto 10cm, 20cm, 30cm. Plantar foot forces and pressures were recorded by the Footmat system(Tekscan, Boston, USA) during level and obstacle walking with barefoot. Plantar foot surface was defined as seven regions for pressure measurement; two toe regions, three forefoot regions, one midfoot region, one heel region. One-way ANOVA was used to compare each region data of foot according to various heights. The results indicated that there are significant differences on peak pressure and maximal forces regarding each region at stance phase. As height of obstacle became high, the pathway of COP had a tendency to be short and abducted. Plantar pressure of foot could be changed by obstacle height and these findings demonstrated that obstacle with different height have an effect on structure and function of the foot.

• Radiation Oncology Journal
• /
• v.10 no.1
• /
• pp.107-113
• /
• 1992

Increasing frequency of skin cancer, mycosis fungoides, Kaposi's sarcoma etc, it need to treatment dose planning for total skin electron beam (TSEB) therapy. Appropriate treatment planning for TSEB therapy is needed to give homogeneous dose distribution throughout the entire skin surface. The energy of 6 MeV electron from the 18 MeV medical linear accelerator was adapted for superficial total skin electron beam therapy. The energy of the electron beam was reduced to 4.2 MeV by a $0.5\;cm\times90\;cm{\times}180\;cm$ acryl screen placed in a feet front of the patient. Six dual field beam was adapted for total skin irradiation to encompass the entire body surface from head to toe simultaneously. The patients were treated behind the acryl screen plate acted as a beam scatterer and contained a parallel-plate shallow ion chamber for dosimetry and beam monitoring. During treatment, the patient was placed in six different positions due to be homogeneous dose distribution for whole skin around the body. One treatment session delivered 400 cGy to the entire skin surface and patients were treated twice a week for eight consecutive weeks, which is equivalent to TDF value 57. instrumentation and techniques developed in determining the depth dose, dose distribution and bremsstrahlung dose are discussed.